Final - Image Stacking - V7.0

Data Load

train_df = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\train\\train.csv')
test_df = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\test\\test.csv')
color_df = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\ColorLabels.csv')
breed_df = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\BreedLabels.csv')

EDA

#Loading libraries
import pandas as pd
import numpy as np

#Libraries for visualization
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline
from sklearn.preprocessing import LabelEncoder

#Loading dataset
df_train = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\train\\train.csv')
df_test = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\test\\test.csv')


#Displaying First 5 rows
df_train.head()

	Type	Name	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	...	Health	Quantity	Fee	State	RescuerID	VideoAmt	Description	PetID	PhotoAmt	AdoptionSpeed
0	2	Nibble	3	299	0	1	1	7	0	1	...	1	1	100	41326	8480853f516546f6cf33aa88cd76c379	0	Nibble is a 3+ month old ball of cuteness. He ...	86e1089a3	1.0	2
1	2	No Name Yet	1	265	0	1	1	2	0	2	...	1	1	0	41401	3082c7125d8fb66f7dd4bff4192c8b14	0	I just found it alone yesterday near my apartm...	6296e909a	2.0	0
2	1	Brisco	1	307	0	1	2	7	0	2	...	1	1	0	41326	fa90fa5b1ee11c86938398b60abc32cb	0	Their pregnant mother was dumped by her irresp...	3422e4906	7.0	3
3	1	Miko	4	307	0	2	1	2	0	2	...	1	1	150	41401	9238e4f44c71a75282e62f7136c6b240	0	Good guard dog, very alert, active, obedience ...	5842f1ff5	8.0	2
4	1	Hunter	1	307	0	1	1	0	0	2	...	1	1	0	41326	95481e953f8aed9ec3d16fc4509537e8	0	This handsome yet cute boy is up for adoption....	850a43f90	3.0	2

5 rows × 24 columns

Data Description

• Overviewing of data types, missing data, and number of rows

df_train.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 14993 entries, 0 to 14992
Data columns (total 24 columns):
 #   Column         Non-Null Count  Dtype  
---  ------         --------------  -----  
 0   Type           14993 non-null  int64  
 1   Name           13728 non-null  object 
 2   Age            14993 non-null  int64  
 3   Breed1         14993 non-null  int64  
 4   Breed2         14993 non-null  int64  
 5   Gender         14993 non-null  int64  
 6   Color1         14993 non-null  int64  
 7   Color2         14993 non-null  int64  
 8   Color3         14993 non-null  int64  
 9   MaturitySize   14993 non-null  int64  
 10  FurLength      14993 non-null  int64  
 11  Vaccinated     14993 non-null  int64  
 12  Dewormed       14993 non-null  int64  
 13  Sterilized     14993 non-null  int64  
 14  Health         14993 non-null  int64  
 15  Quantity       14993 non-null  int64  
 16  Fee            14993 non-null  int64  
 17  State          14993 non-null  int64  
 18  RescuerID      14993 non-null  object 
 19  VideoAmt       14993 non-null  int64  
 20  Description    14980 non-null  object 
 21  PetID          14993 non-null  object 
 22  PhotoAmt       14993 non-null  float64
 23  AdoptionSpeed  14993 non-null  int64  
dtypes: float64(1), int64(19), object(4)
memory usage: 2.7+ MB

• The distribution of numerical and categorical data

df_train.describe().T

	count	mean	std	min	25%	50%	75%	max
Type	14993.0	1.457614	0.498217	1.0	1.0	1.0	2.0	2.0
Age	14993.0	10.452078	18.155790	0.0	2.0	3.0	12.0	255.0
Breed1	14993.0	265.272594	60.056818	0.0	265.0	266.0	307.0	307.0
Breed2	14993.0	74.009738	123.011575	0.0	0.0	0.0	179.0	307.0
Gender	14993.0	1.776162	0.681592	1.0	1.0	2.0	2.0	3.0
Color1	14993.0	2.234176	1.745225	1.0	1.0	2.0	3.0	7.0
Color2	14993.0	3.222837	2.742562	0.0	0.0	2.0	6.0	7.0
Color3	14993.0	1.882012	2.984086	0.0	0.0	0.0	5.0	7.0
MaturitySize	14993.0	1.862002	0.547959	1.0	2.0	2.0	2.0	4.0
FurLength	14993.0	1.467485	0.599070	1.0	1.0	1.0	2.0	3.0
Vaccinated	14993.0	1.731208	0.667649	1.0	1.0	2.0	2.0	3.0
Dewormed	14993.0	1.558727	0.695817	1.0	1.0	1.0	2.0	3.0
Sterilized	14993.0	1.914227	0.566172	1.0	2.0	2.0	2.0	3.0
Health	14993.0	1.036617	0.199535	1.0	1.0	1.0	1.0	3.0
Quantity	14993.0	1.576069	1.472477	1.0	1.0	1.0	1.0	20.0
Fee	14993.0	21.259988	78.414548	0.0	0.0	0.0	0.0	3000.0
State	14993.0	41346.028347	32.444153	41324.0	41326.0	41326.0	41401.0	41415.0
VideoAmt	14993.0	0.056760	0.346185	0.0	0.0	0.0	0.0	8.0
PhotoAmt	14993.0	3.889215	3.487810	0.0	2.0	3.0	5.0	30.0
AdoptionSpeed	14993.0	2.516441	1.177265	0.0	2.0	2.0	4.0	4.0

df_test.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3972 entries, 0 to 3971
Data columns (total 23 columns):
 #   Column        Non-Null Count  Dtype  
---  ------        --------------  -----  
 0   Type          3972 non-null   int64  
 1   Name          3558 non-null   object 
 2   Age           3972 non-null   int64  
 3   Breed1        3972 non-null   int64  
 4   Breed2        3972 non-null   int64  
 5   Gender        3972 non-null   int64  
 6   Color1        3972 non-null   int64  
 7   Color2        3972 non-null   int64  
 8   Color3        3972 non-null   int64  
 9   MaturitySize  3972 non-null   int64  
 10  FurLength     3972 non-null   int64  
 11  Vaccinated    3972 non-null   int64  
 12  Dewormed      3972 non-null   int64  
 13  Sterilized    3972 non-null   int64  
 14  Health        3972 non-null   int64  
 15  Quantity      3972 non-null   int64  
 16  Fee           3972 non-null   int64  
 17  State         3972 non-null   int64  
 18  RescuerID     3972 non-null   object 
 19  VideoAmt      3972 non-null   int64  
 20  Description   3971 non-null   object 
 21  PetID         3972 non-null   object 
 22  PhotoAmt      3972 non-null   float64
dtypes: float64(1), int64(18), object(4)
memory usage: 713.8+ KB

df_train.columns

Index(['Type', 'Name', 'Age', 'Breed1', 'Breed2', 'Gender', 'Color1', 'Color2',
       'Color3', 'MaturitySize', 'FurLength', 'Vaccinated', 'Dewormed',
       'Sterilized', 'Health', 'Quantity', 'Fee', 'State', 'RescuerID',
       'VideoAmt', 'Description', 'PetID', 'PhotoAmt', 'AdoptionSpeed'],
      dtype='object')

• Data inspection
• Handling missing values
• Outlier detection
• Data transformation

Data Inspection

#Data Inspection

#Looking for missing values
df_train.isna()

	Type	Name	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	...	Health	Quantity	Fee	State	RescuerID	VideoAmt	Description	PetID	PhotoAmt	AdoptionSpeed
0	False	False	False	False	False	False	False	False	False	False	...	False	False	False	False	False	False	False	False	False	False
1	False	False	False	False	False	False	False	False	False	False	...	False	False	False	False	False	False	False	False	False	False
2	False	False	False	False	False	False	False	False	False	False	...	False	False	False	False	False	False	False	False	False	False
3	False	False	False	False	False	False	False	False	False	False	...	False	False	False	False	False	False	False	False	False	False
4	False	False	False	False	False	False	False	False	False	False	...	False	False	False	False	False	False	False	False	False	False
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
14988	False	True	False	False	False	False	False	False	False	False	...	False	False	False	False	False	False	False	False	False	False
14989	False	False	False	False	False	False	False	False	False	False	...	False	False	False	False	False	False	False	False	False	False
14990	False	False	False	False	False	False	False	False	False	False	...	False	False	False	False	False	False	False	False	False	False
14991	False	False	False	False	False	False	False	False	False	False	...	False	False	False	False	False	False	False	False	False	False
14992	False	False	False	False	False	False	False	False	False	False	...	False	False	False	False	False	False	False	False	False	False

14993 rows × 24 columns

#Checking for columns with missing values
df_train.isnull().sum()

Type                0
Name             1265
Age                 0
Breed1              0
Breed2              0
Gender              0
Color1              0
Color2              0
Color3              0
MaturitySize        0
FurLength           0
Vaccinated          0
Dewormed            0
Sterilized          0
Health              0
Quantity            0
Fee                 0
State               0
RescuerID           0
VideoAmt            0
Description        13
PetID               0
PhotoAmt            0
AdoptionSpeed       0
dtype: int64

• Treating missing values for Names, checking and removing for duplicates

#Filling missing values with Unknown

df_train['Name'].fillna("Unknown", inplace = True)
df_train['Description'].fillna("Unknown", inplace = True)

df_test['Name'].fillna("Unknown", inplace = True)
df_test['Description'].fillna("Unknown", inplace = True)

#Filling missing numeric values with median and mean

df_train['Age'].replace(0, df_train['Age'].median(), inplace = True)
df_train['Fee'].replace(0, df_train['Fee'].mean(), inplace = True)

df_test['Age'].replace(0, df_test['Age'].median(), inplace = True)
df_test['Fee'].replace(0, df_test['Fee'].mean(), inplace = True)

#Checking for and remove duplicate rows
df_train = df_train.drop_duplicates()
df_test = df_test.drop_duplicates()

df_train.info()
print('_'*40)
df_test.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 14993 entries, 0 to 14992
Data columns (total 24 columns):
 #   Column         Non-Null Count  Dtype  
---  ------         --------------  -----  
 0   Type           14993 non-null  int64  
 1   Name           14993 non-null  object 
 2   Age            14993 non-null  int64  
 3   Breed1         14993 non-null  int64  
 4   Breed2         14993 non-null  int64  
 5   Gender         14993 non-null  int64  
 6   Color1         14993 non-null  int64  
 7   Color2         14993 non-null  int64  
 8   Color3         14993 non-null  int64  
 9   MaturitySize   14993 non-null  int64  
 10  FurLength      14993 non-null  int64  
 11  Vaccinated     14993 non-null  int64  
 12  Dewormed       14993 non-null  int64  
 13  Sterilized     14993 non-null  int64  
 14  Health         14993 non-null  int64  
 15  Quantity       14993 non-null  int64  
 16  Fee            14993 non-null  float64
 17  State          14993 non-null  int64  
 18  RescuerID      14993 non-null  object 
 19  VideoAmt       14993 non-null  int64  
 20  Description    14993 non-null  object 
 21  PetID          14993 non-null  object 
 22  PhotoAmt       14993 non-null  float64
 23  AdoptionSpeed  14993 non-null  int64  
dtypes: float64(2), int64(18), object(4)
memory usage: 2.7+ MB
________________________________________
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3972 entries, 0 to 3971
Data columns (total 23 columns):
 #   Column        Non-Null Count  Dtype  
---  ------        --------------  -----  
 0   Type          3972 non-null   int64  
 1   Name          3972 non-null   object 
 2   Age           3972 non-null   int64  
 3   Breed1        3972 non-null   int64  
 4   Breed2        3972 non-null   int64  
 5   Gender        3972 non-null   int64  
 6   Color1        3972 non-null   int64  
 7   Color2        3972 non-null   int64  
 8   Color3        3972 non-null   int64  
 9   MaturitySize  3972 non-null   int64  
 10  FurLength     3972 non-null   int64  
 11  Vaccinated    3972 non-null   int64  
 12  Dewormed      3972 non-null   int64  
 13  Sterilized    3972 non-null   int64  
 14  Health        3972 non-null   int64  
 15  Quantity      3972 non-null   int64  
 16  Fee           3972 non-null   float64
 17  State         3972 non-null   int64  
 18  RescuerID     3972 non-null   object 
 19  VideoAmt      3972 non-null   int64  
 20  Description   3972 non-null   object 
 21  PetID         3972 non-null   object 
 22  PhotoAmt      3972 non-null   float64
dtypes: float64(2), int64(17), object(4)
memory usage: 713.8+ KB

#Checking for missing values again
df_train.isnull().sum()

Type             0
Name             0
Age              0
Breed1           0
Breed2           0
Gender           0
Color1           0
Color2           0
Color3           0
MaturitySize     0
FurLength        0
Vaccinated       0
Dewormed         0
Sterilized       0
Health           0
Quantity         0
Fee              0
State            0
RescuerID        0
VideoAmt         0
Description      0
PetID            0
PhotoAmt         0
AdoptionSpeed    0
dtype: int64

• Outliers Detection

sns.set(style="whitegrid")
plt.figure(figsize=(10,5))
ax = sns.boxplot(x='Age', data=df_train, orient="h")

print(np.where(df_train['Age']>100))

(array([  113,   145,   513,  1057,  1141,  1363,  1703,  1858,  1867,
        2263,  2497,  2767,  2830,  3195,  3684,  3931,  3998,  4071,
        4344,  4456,  4522,  4617,  4796,  5007,  5067,  5160,  5485,
        5535,  5569,  5593,  6264,  6569,  6618,  6666,  6722,  6894,
        6992,  7166,  7250,  7289,  7482,  7512,  7614,  7823,  8058,
        8065,  8304,  8492,  8944,  9267,  9378,  9563,  9970, 10089,
       10128, 10337, 10426, 10501, 10658, 10737, 10917, 11087, 11172,
       11216, 11314, 11523, 11605, 11640, 11968, 12526, 12575, 12621,
       12737, 12855, 12947, 13079, 13398, 13493, 13665, 13744, 14189,
       14303, 14637, 14732], dtype=int64),)

sns.set(style="whitegrid")
plt.figure(figsize=(10,5))
ax = sns.boxplot(x='Fee', data=df_train, orient="h")

print(np.where(df_train['Fee']>500))

(array([   81,   901,   953,  1459,  2038,  2078,  2247,  2519,  2913,
        3701,  4356,  4797,  4844,  5092,  5289,  5902,  6077,  6306,
        7315,  8590,  8722,  8834,  8879,  9003,  9393,  9707,  9745,
        9782,  9947, 10477, 10621, 10624, 10930, 11426, 11519, 11687,
       12304, 12365, 12582, 13121, 13631, 14245, 14454, 14519],
      dtype=int64),)

sns.set(style="whitegrid")
plt.figure(figsize=(10,5))
ax = sns.boxplot(x='Quantity', data=df_train, orient="h")

print(np.where(df_train['Quantity']>10))

(array([  263,   320,   396,   427,   823,  1088,  1302,  1305,  1377,
        1404,  1659,  2558,  2787,  2886,  3894,  4759,  5209,  5227,
        6091,  6291,  6332,  6997,  7094,  7847,  7875,  7966,  7993,
        8424,  8821,  9306,  9405,  9779,  9857,  9926, 11101, 11568,
       11729, 12792, 13383, 13596, 13599, 13865, 14583], dtype=int64),)

Data Preprocessing

*Data normalizarion, scaling and transformation

# Using standardization for scaling train data outliers

# For numeric columns
numeric_cols = ['Age', 'Fee', 'Quantity']

for column in numeric_cols:
  df_train[numeric_cols] = (df_train[numeric_cols] - df_train[numeric_cols].mean()) / df_train[numeric_cols].std()

df_train.tail(2)

	Type	Name	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	...	Health	Quantity	Fee	State	RescuerID	VideoAmt	Description	PetID	PhotoAmt	AdoptionSpeed
14991	2	Ms Daym	-0.082032	266	0	2	4	7	0	1	...	1	-0.391224	-0.243348	41336	c311c0c569245baa147d91fa4e351ae4	0	she is very shy..adventures and independent..s...	e4da1c9e4	3.0	4
14992	1	Fili	-0.523092	307	307	1	2	0	0	2	...	1	-0.391224	-0.243348	41332	9ed1d5493d223eaa5024c1a031dbc9c2	0	Fili just loves laying around and also loves b...	a83d95ead	1.0	3

2 rows × 24 columns

# Using standardization for scaling test data outliers

# For numeric columns
for column in numeric_cols:
  df_test[numeric_cols] = (df_test[numeric_cols] - df_test[numeric_cols].mean()) / df_test[numeric_cols].std()

df_test.tail(2)

	Type	Name	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	...	Sterilized	Health	Quantity	Fee	State	RescuerID	VideoAmt	Description	PetID	PhotoAmt
3970	1	Smokey	0.725031	307	0	2	5	7	0	2	...	2	1	-0.397970	-0.25028	41325	087903c2819a6297519c93d962b488d5	0	smokey is good family pet. very obedient,so lo...	fd80b8c80	3.0
3971	2	Unknown	-0.591002	266	0	3	1	7	0	1	...	2	1	2.934107	-0.25028	41326	041876679ae0261e3ee4aefe9e66316f	0	Found abandoned with the mother at nearby shop...	493ed84ae	1.0

2 rows × 23 columns

• Exploring the correlation within features and the target variable:

• Upon analyzing the frequency counts for pet types (1 for 'Dog' and 2 for 'Cat') and Adoption Speed (target variable).

• The Adoption Speed is a categorical feature that represents the speed of adoption. For instance, lower is faster.

• We should use Chi-squared test for the association between each categorical variable and the target variable

plt.figure(figsize=(12,8))
axis = sns.countplot(x = "Type", hue = "AdoptionSpeed", palette = "Set1", data = df_train)
axis.set(title = "No. of Pets adopted by Type", xlabel = "Type",ylabel = "No.of Pets")
plt.legend(loc='upper right', title='Adoption Speed')
plt.show()

It's observed there are slightly more Dogs than Cats and the Cats are adopted a bit faster.

df_dogs = df_train[df_train['Type']==1]
df_dogs[['AdoptionSpeed', 'PetID']].groupby(['AdoptionSpeed'], as_index=False).count().sort_values(by='AdoptionSpeed', ascending=True)

	AdoptionSpeed	PetID
0	0	170
1	1	1435
2	2	2164
3	3	1949
4	4	2414

df_cats = df_train[df_train['Type']==2]
df_cats[['AdoptionSpeed', 'PetID']].groupby(['AdoptionSpeed'], as_index=False).count().sort_values(by='AdoptionSpeed', ascending=True)

	AdoptionSpeed	PetID
0	0	240
1	1	1655
2	2	1873
3	3	1310
4	4	1783

The following feature provides the gender of pet. It is coded as follows: 1 for 'Male', 2 for 'Female', and 3 for 'Mixed' when the profile represents a group of pets.

plt.figure(figsize=(12,8))
axis = sns.countplot(x = "Gender", hue = "AdoptionSpeed", palette = "Set1", data = df_dogs )
axis.set(title = "The Adoption Speed by Gender", xlabel = "Gender",ylabel = "No.of Dogs")
plt.show()

plt.figure(figsize=(12,8))
axis = sns.countplot(x = "Gender", hue = "AdoptionSpeed", palette = "Set1", data = df_cats )
axis.set(title = "The Adoption Speed by Gender", xlabel = "Gender",ylabel = "No.of Cats")
plt.show()

# Adding Gender labels
df_train['GenderName'] = df_train['Gender'].map({1:'Male', 2:'Female', 3:'Mixed'})

grid = sns.FacetGrid(df_train, col='GenderName',row='Type', aspect=1.6)
grid.map(plt.hist, 'Age', alpha=.75, bins=12)
grid.add_legend()

<seaborn.axisgrid.FacetGrid at 0x1e60b379790>

Examining the chart above, we can observe a broad distribution of age among female cats and dogs.

The pets exhibited a high distribution in age, indicating they were recently born.

It also indicates that pets with mixed genders had the lowest age distribution.

grid = sns.FacetGrid(df_train, col='GenderName',row='AdoptionSpeed', aspect=1.6)
grid.map(plt.hist, 'Age', alpha=.8, bins=10)
grid.add_legend()

<seaborn.axisgrid.FacetGrid at 0x1e60e4839e0>

numeric_cols = df_train.select_dtypes(include=[np.number])
correlation_matrix = numeric_cols.corr()
numeric_cols.corr()

	Type	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	FurLength	Vaccinated	Dewormed	Sterilized	Health	Quantity	Fee	State	VideoAmt	PhotoAmt	AdoptionSpeed
Type	1.000000	-0.147038	0.057724	-0.043568	0.060843	0.097680	0.246333	0.201693	-0.171811	0.003036	0.102907	0.025508	0.006737	-0.006864	0.036423	-0.047583	0.124549	-0.005967	0.049368	-0.091240
Age	-0.147038	1.000000	-0.314597	-0.041049	-0.122215	0.089955	-0.039586	-0.052880	0.093094	0.152823	-0.135151	-0.051985	-0.188966	0.103728	-0.111202	0.089362	0.024514	-0.016085	-0.084709	0.100126
Breed1	0.057724	-0.314597	1.000000	-0.157053	0.071321	-0.035510	-0.007081	-0.000682	-0.010507	-0.108059	0.046518	0.013216	0.052698	-0.034893	0.090194	-0.190975	-0.026832	0.020626	0.043395	0.107834
Breed2	-0.043568	-0.041049	-0.157053	1.000000	0.056907	-0.020324	0.003015	0.035679	0.049879	0.106220	0.005638	-0.009817	-0.005989	-0.029285	0.038125	0.010657	-0.037993	0.001569	0.045953	-0.018642
Gender	0.060843	-0.122215	0.071321	0.056907	1.000000	-0.115351	0.026507	0.259309	-0.091819	-0.030404	0.078702	0.091431	0.040645	-0.045177	0.494489	-0.047893	0.002564	0.016816	0.098435	0.057622
Color1	0.097680	0.089955	-0.035510	-0.020324	-0.115351	1.000000	-0.113290	-0.282897	-0.029468	0.066967	-0.017475	-0.019318	-0.037050	0.024410	-0.116688	0.050779	0.024012	-0.006435	-0.041323	-0.044192
Color2	0.246333	-0.039586	-0.007081	0.003015	0.026507	-0.113290	1.000000	0.088742	-0.065155	-0.008764	0.026193	0.005812	0.007370	-0.004307	0.019876	-0.018441	0.026944	0.023983	0.059113	-0.038579
Color3	0.201693	-0.052880	-0.000682	0.035679	0.259309	-0.282897	0.088742	1.000000	-0.051730	0.013581	0.054053	0.048617	0.038899	-0.024782	0.272548	-0.020458	0.010554	0.019526	0.103662	-0.007014
MaturitySize	-0.171811	0.093094	-0.010507	0.049879	-0.091819	-0.029468	-0.065155	-0.051730	1.000000	0.095142	-0.087722	-0.067352	-0.066321	-0.012957	-0.042832	0.040549	-0.062584	0.017032	0.015593	0.045653
FurLength	0.003036	0.152823	-0.108059	0.106220	-0.030404	0.066967	-0.008764	0.013581	0.095142	1.000000	-0.006010	0.016463	0.032092	0.028096	-0.038388	0.158679	-0.029317	-0.014740	-0.031557	-0.091337
Vaccinated	0.102907	-0.135151	0.046518	0.005638	0.078702	-0.017475	0.026193	0.054053	-0.087722	-0.006010	1.000000	0.722596	0.471204	0.078893	0.125287	-0.117371	0.032737	-0.032685	-0.052719	-0.059043
Dewormed	0.025508	-0.051985	0.013216	-0.009817	0.091431	-0.019318	0.005812	0.048617	-0.067352	0.016463	0.722596	1.000000	0.426256	0.072188	0.141355	-0.100696	-0.001310	-0.036961	-0.099026	-0.013210
Sterilized	0.006737	-0.188966	0.052698	-0.005989	0.040645	-0.037050	0.007370	0.038899	-0.066321	0.032092	0.471204	0.426256	1.000000	0.057916	0.101919	-0.059366	0.009951	-0.018720	-0.062067	-0.083447
Health	-0.006864	0.103728	-0.034893	-0.029285	-0.045177	0.024410	-0.004307	-0.024782	-0.012957	0.028096	0.078893	0.072188	0.057916	1.000000	-0.034567	-0.009325	0.026237	-0.004984	-0.023978	0.029382
Quantity	0.036423	-0.111202	0.090194	0.038125	0.494489	-0.116688	0.019876	0.272548	-0.042832	-0.038388	0.125287	0.141355	0.101919	-0.034567	1.000000	-0.057984	-0.002182	0.008344	0.137839	0.062931
Fee	-0.047583	0.089362	-0.190975	0.010657	-0.047893	0.050779	-0.018441	-0.020458	0.040549	0.158679	-0.117371	-0.100696	-0.059366	-0.009325	-0.057984	1.000000	-0.012178	-0.001377	0.005775	-0.006288
State	0.124549	0.024514	-0.026832	-0.037993	0.002564	0.024012	0.026944	0.010554	-0.062584	-0.029317	0.032737	-0.001310	0.009951	0.026237	-0.002182	-0.012178	1.000000	-0.027456	0.001873	0.013119
VideoAmt	-0.005967	-0.016085	0.020626	0.001569	0.016816	-0.006435	0.023983	0.019526	0.017032	-0.014740	-0.032685	-0.036961	-0.018720	-0.004984	0.008344	-0.001377	-0.027456	1.000000	0.227231	-0.000571
PhotoAmt	0.049368	-0.084709	0.043395	0.045953	0.098435	-0.041323	0.059113	0.103662	0.015593	-0.031557	-0.052719	-0.099026	-0.062067	-0.023978	0.137839	0.005775	0.001873	0.227231	1.000000	-0.023233
AdoptionSpeed	-0.091240	0.100126	0.107834	-0.018642	0.057622	-0.044192	-0.038579	-0.007014	0.045653	-0.091337	-0.059043	-0.013210	-0.083447	0.029382	0.062931	-0.006288	0.013119	-0.000571	-0.023233	1.000000

# Compute the correlation between features and the target variable (e.g., 'AdoptionSpeed')
corr_to_target = numeric_cols.corr()['AdoptionSpeed'].abs().sort_values(ascending=False)

# Display the top N features with the highest absolute correlation to the target variable
top_n = 19 # Change this number to display a different number of top features
top_features_to_target = corr_to_target[1:top_n]  # Exclude the target variable
print("Top", top_n, "features with highest correlation to the target:")
print(top_features_to_target)

Top 19 features with highest correlation to the target:
Breed1          0.107834
Age             0.100126
FurLength       0.091337
Type            0.091240
Sterilized      0.083447
Quantity        0.062931
Vaccinated      0.059043
Gender          0.057622
MaturitySize    0.045653
Color1          0.044192
Color2          0.038579
Health          0.029382
PhotoAmt        0.023233
Breed2          0.018642
Dewormed        0.013210
State           0.013119
Color3          0.007014
Fee             0.006288
Name: AdoptionSpeed, dtype: float64

# Compute the correlation matrix between all features
correlation_matrix_1 = correlation_matrix.corr(method='spearman').round(2)

plt.figure(figsize=(17,10))
sns.heatmap(correlation_matrix, cmap='coolwarm', linewidth = 0.5, annot=True)
plt.title('Correlation Heatmap')
plt.show()

# Specify the correlation threshold as values greater than 0.4 or less than -0.4
correlation_threshold = 0.4  # We can adjust this value

# Create a mask to identify pairs of features with correlation exceeding the threshold
high_correlation_mask = (abs(correlation_matrix) > correlation_threshold) & (correlation_matrix != 1.0)

# Identify the indices (pairs) of highly correlated features
high_correlation_pairs = [(i, j) for i in range(len(correlation_matrix.columns)) for j in range(i + 1, len(correlation_matrix.columns)) if high_correlation_mask.iloc[i, j]]

# Determine which feature to drop from each high-correlation pair
features_to_drop = []
for i, j in high_correlation_pairs:
    feature_i = correlation_matrix.columns[i]
    feature_j = correlation_matrix.columns[j]

    # Drop the feature with the higher index
    if i > j:
        features_to_drop.append(feature_i)
    else:
        features_to_drop.append(feature_j)

features_to_drop = set(features_to_drop)
features_to_drop

{'Dewormed', 'Quantity', 'Sterilized'}

Feature Engineering

import os

• Tabular Data Features

train_df = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\train\\train.csv')
test_df = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\test\\test.csv')
color_df = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\ColorLabels.csv')
breed_df = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\BreedLabels.csv')

output_dir = 'C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output'

train_df.head()

	Type	Name	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	...	Health	Quantity	Fee	State	RescuerID	VideoAmt	Description	PetID	PhotoAmt	AdoptionSpeed
0	2	Nibble	3	299	0	1	1	7	0	1	...	1	1	100	41326	8480853f516546f6cf33aa88cd76c379	0	Nibble is a 3+ month old ball of cuteness. He ...	86e1089a3	1.0	2
1	2	No Name Yet	1	265	0	1	1	2	0	2	...	1	1	0	41401	3082c7125d8fb66f7dd4bff4192c8b14	0	I just found it alone yesterday near my apartm...	6296e909a	2.0	0
2	1	Brisco	1	307	0	1	2	7	0	2	...	1	1	0	41326	fa90fa5b1ee11c86938398b60abc32cb	0	Their pregnant mother was dumped by her irresp...	3422e4906	7.0	3
3	1	Miko	4	307	0	2	1	2	0	2	...	1	1	150	41401	9238e4f44c71a75282e62f7136c6b240	0	Good guard dog, very alert, active, obedience ...	5842f1ff5	8.0	2
4	1	Hunter	1	307	0	1	1	0	0	2	...	1	1	0	41326	95481e953f8aed9ec3d16fc4509537e8	0	This handsome yet cute boy is up for adoption....	850a43f90	3.0	2

5 rows × 24 columns

# basic features : if its a mix of both breed, it will be 2 else 1.
train_df['num_breed'] = ((train_df['Breed1'] > 0) & (train_df['Breed2'] > 0)).astype(int) + ((train_df['Breed1'] > 0) | (train_df['Breed2'] > 0)).astype(int)
test_df['num_breed'] = ((test_df['Breed1'] > 0) & (test_df['Breed2'] > 0)).astype(int) + ((test_df['Breed1'] > 0) | (test_df['Breed2'] > 0)).astype(int)

# same with number of colors
color_columns = ['Color1', 'Color2', 'Color3']
train_df['num_colors'] = (train_df[color_columns] > 0).sum(axis=1)
test_df['num_colors'] = (test_df[color_columns] > 0).sum(axis=1)

# lets do something with fee
import matplotlib.pyplot as plt
# distribution of fee
plt.hist(train_df['Fee'], bins=100, color='blue', edgecolor='black')
plt.title('Distribution of Fee')
plt.xlabel('Fee')
plt.ylabel('Frequency')
plt.xlim(0,300)
plt.show()

# using vaccinated, dewormed, health and sterilized features
train_df['Total_Vaccinations'] = train_df[['Vaccinated', 'Dewormed', 'Sterilized']].sum(axis=1)
test_df['Total_Vaccinations'] = test_df[['Vaccinated', 'Dewormed', 'Sterilized']].sum(axis=1)

train_df['Avg_Health_Score'] = train_df[['Health']].mean(axis=1)
test_df['Avg_Health_Score'] = test_df[['Health']].mean(axis=1)

# feature using fur length and health

from sklearn.preprocessing import LabelEncoder
label_encoder = LabelEncoder()

# creating interaction feature
train_df['Fur_Health_Interact'] = train_df['FurLength'].astype(str) + '_' + train_df['Health'].astype(str)
# Fit and transform the column
train_df['Fur_Health_Interact_Encoded'] = label_encoder.fit_transform(train_df['Fur_Health_Interact'])

test_df['Fur_Health_Interact'] = test_df['FurLength'].astype(str) + '_' + test_df['Health'].astype(str)
# Fit and transform the column
test_df['Fur_Health_Interact_Encoded'] = label_encoder.fit_transform(test_df['Fur_Health_Interact'])

# feature using health
def feature_eng_health_vaccinated(df):

    conditions = [
        (df['Health'] == 1) & (df['Total_Vaccinations'] == 0),
        (df['Health'] == 1) & (df['Total_Vaccinations'] > 0),
        (df['Health'].isin([2, 3]))
    ]
    choices = ['Healthy and Unvaccinated', 'Healthy and Vaccinated', 'Injured or Sick']
    df['Overall_Health_Status'] = np.select(conditions, choices, default='Not Specified')
    df['Overall_Health_Status_Encoded'] = label_encoder.fit_transform(df['Overall_Health_Status'])

    return df

train_df = feature_eng_health_vaccinated(train_df)
test_df = feature_eng_health_vaccinated(test_df)

# arranging the column order
new_column_order = ['PetID'] + [col for col in train_df.columns if col != 'PetID' and col != 'AdoptionSpeed'] + ['AdoptionSpeed']
train_df_new_order = train_df[new_column_order]

if 'AdoptionSpeed' in test_df.columns:
    new_column_order = ['PetID'] + [col for col in test_df.columns if col != 'PetID' and col != 'AdoptionSpeed'] + ['AdoptionSpeed']
    test_df = test_df[new_column_order]
else:
    new_column_order = ['PetID'] + [col for col in test_df.columns if col != 'PetID']
    test_df = test_df[new_column_order]

test_df.shape

(3972, 31)

# saving the files
train_df_new_order.to_csv(os.path.join(output_dir, 'train_df_mahmood_features.csv'), index=False)
test_df.to_csv(os.path.join(output_dir, 'test_df_mahmood_features.csv'), index=False)

• Sentiment Data

import pandas as pd
import nltk
from nltk.sentiment import SentimentIntensityAnalyzer
from textblob import TextBlob

file_path = 'C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\train\\train.csv'
file_path_test = 'C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\test\\test.csv'

data = pd.read_csv(file_path)

data.head()

	Type	Name	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	...	Health	Quantity	Fee	State	RescuerID	VideoAmt	Description	PetID	PhotoAmt	AdoptionSpeed
0	2	Nibble	3	299	0	1	1	7	0	1	...	1	1	100	41326	8480853f516546f6cf33aa88cd76c379	0	Nibble is a 3+ month old ball of cuteness. He ...	86e1089a3	1.0	2
1	2	No Name Yet	1	265	0	1	1	2	0	2	...	1	1	0	41401	3082c7125d8fb66f7dd4bff4192c8b14	0	I just found it alone yesterday near my apartm...	6296e909a	2.0	0
2	1	Brisco	1	307	0	1	2	7	0	2	...	1	1	0	41326	fa90fa5b1ee11c86938398b60abc32cb	0	Their pregnant mother was dumped by her irresp...	3422e4906	7.0	3
3	1	Miko	4	307	0	2	1	2	0	2	...	1	1	150	41401	9238e4f44c71a75282e62f7136c6b240	0	Good guard dog, very alert, active, obedience ...	5842f1ff5	8.0	2
4	1	Hunter	1	307	0	1	1	0	0	2	...	1	1	0	41326	95481e953f8aed9ec3d16fc4509537e8	0	This handsome yet cute boy is up for adoption....	850a43f90	3.0	2

5 rows × 24 columns

data_test = pd.read_csv(file_path_test)

data_test.head()

	Type	Name	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	...	Sterilized	Health	Quantity	Fee	State	RescuerID	VideoAmt	Description	PetID	PhotoAmt
0	2	Dopey & Grey	8	266	266	1	2	6	7	1	...	2	1	2	0	41326	2ece3b2573dcdcebd774e635dca15fd9	0	Dopey Age: 8mths old Male One half of a pair, ...	e2dfc2935	2.0
1	2	Chi Chi	36	285	264	2	1	4	7	2	...	1	2	1	0	41326	2ece3b2573dcdcebd774e635dca15fd9	0	Please note that Chichi has been neutered, the...	f153b465f	1.0
2	2	Sticky	2	265	0	1	6	7	0	2	...	2	1	1	200	41326	e59c106e9912fa30c898976278c2e834	0	Sticky, named such because of his tendency to ...	3c90f3f54	4.0
3	1	Dannie & Kass [In Penang]	12	307	0	2	2	5	0	2	...	1	1	2	0	41326	e59c106e9912fa30c898976278c2e834	0	Dannie and Kass are mother and daughter. We en...	e02abc8a3	5.0
4	2	Cuddles	12	265	0	1	2	3	7	2	...	1	1	1	0	41326	e59c106e9912fa30c898976278c2e834	0	Extremely cuddly cat, hence the origin of his ...	09f0df7d1	5.0

5 rows × 23 columns

data['Description'] = data['Description'].astype(str).fillna("No description")

nltk.download('vader_lexicon')

[nltk_data] Downloading package vader_lexicon to
[nltk_data]     C:\Users\leoko\AppData\Roaming\nltk_data...
[nltk_data]   Package vader_lexicon is already up-to-date!

True

sia = SentimentIntensityAnalyzer()

data['Description'] = data['Description'].astype(str).fillna("No description")

data['Description'] = data['Description'].astype(str).fillna('')

data['sentiments'] = data['Description'].apply(lambda x: sia.polarity_scores(x))

data['compound'] = data['sentiments'].apply(lambda x: x['compound'])

print(data[['Description', 'compound']].head())

                                         Description  compound
0  Nibble is a 3+ month old ball of cuteness. He ...    0.9552
1  I just found it alone yesterday near my apartm...    0.1280
2  Their pregnant mother was dumped by her irresp...    0.7650
3  Good guard dog, very alert, active, obedience ...    0.9538
4  This handsome yet cute boy is up for adoption....    0.9880

data_test['Description'] = data_test['Description'].astype(str).fillna("No description")

data_test['Description'] = data_test['Description'].astype(str).fillna('')

data_test['sentiments'] = data_test['Description'].apply(lambda x: sia.polarity_scores(x))

data_test['compound'] = data_test['sentiments'].apply(lambda x: x['compound'])

print(data_test[['Description', 'compound']].head())

                                         Description  compound
0  Dopey Age: 8mths old Male One half of a pair, ...    0.9696
1  Please note that Chichi has been neutered, the...    0.9446
2  Sticky, named such because of his tendency to ...    0.9758
3  Dannie and Kass are mother and daughter. We en...    0.9792
4  Extremely cuddly cat, hence the origin of his ...    0.8809

train_file_path = 'C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\train\\train.csv'
test_file_path = 'C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\test\\test.csv'

train = pd.read_csv(train_file_path)
test = pd.read_csv(test_file_path)

def get_polarity(text):
    return TextBlob(text).sentiment.polarity

def get_subjectivity(text):
    return TextBlob(text).sentiment.subjectivity

train['polarity'] = train['Description'].astype(str).apply(get_polarity)
train['subjectivity'] = train['Description'].astype(str).apply(get_subjectivity)

test['polarity'] = test['Description'].astype(str).apply(get_polarity)
test['subjectivity'] = test['Description'].astype(str).apply(get_subjectivity)

print(train[['Description', 'polarity', 'subjectivity']].head())
print(test[['Description', 'polarity', 'subjectivity']].head())

                                         Description  polarity  subjectivity
0  Nibble is a 3+ month old ball of cuteness. He ...  0.130357      0.407143
1  I just found it alone yesterday near my apartm...  0.100000      0.400000
2  Their pregnant mother was dumped by her irresp...  0.261905      0.542857
3  Good guard dog, very alert, active, obedience ...  0.341146      0.437500
4  This handsome yet cute boy is up for adoption....  0.400000      0.743750
                                         Description  polarity  subjectivity
0  Dopey Age: 8mths old Male One half of a pair, ...  0.035937      0.230729
1  Please note that Chichi has been neutered, the...  0.054464      0.167445
2  Sticky, named such because of his tendency to ...  0.317857      0.678571
3  Dannie and Kass are mother and daughter. We en...  0.347865      0.603333
4  Extremely cuddly cat, hence the origin of his ...  0.142347      0.657143

# saving the files
train.to_csv(os.path.join(output_dir, 'processed_train.csv'), index=False)
test.to_csv(os.path.join(output_dir, 'processed_test.csv'), index=False)

• Image Features

We are extracting Texture Features and Local Features here. Below are some information about those features :

Haralick Texture Features: These include energy, contrast, correlation, and entropy, which describe the texture of an image.

Local Features:SIFT (Scale-Invariant Feature Transform) and SURF (Speeded-Up Robust Features): Detect and describe distinctive local features in images.

import numpy as np # linear algebra
import pandas as pd
pd.set_option('display.max_rows', None)  # Display all rows
pd.set_option('display.max_columns', None)  # Display all columns
import json
import cv2
import numpy as np
import skimage.feature as skf
import skimage.filters as skg
from skimage.measure import shannon_entropy
from scipy.stats import entropy
from sklearn.cluster import KMeans

Extracting texture features

def extract_texture_features(image_path):
    # Load the image
    img = cv2.imread(image_path)

    # Convert the image to grayscale
    img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

    haralick = skf.graycomatrix(img_gray, [1], [0], 256, symmetric=True, normed=True)
    energy = skf.graycoprops(haralick, 'energy')[0, 0]
    contrast = skf.graycoprops(haralick, 'contrast')[0, 0]
    correlation = skf.graycoprops(haralick, 'correlation')[0, 0]
    entropy = shannon_entropy(img_gray)

    mean = np.mean(img)
    std_dev = np.std(img)

    return energy, contrast, correlation, entropy, mean, std_dev

test_img_path = 'C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\test_images'
train_img_path = 'C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\train_images'


# looping through the training images
# Create an empty dataframe
df_train_img = pd.DataFrame(columns=['PetID', 'Energy', 'Contrast', 'Correlation', 'Entropy', 'Mean', 'Std_Dev'])

# Loop through each image in the folder
for filename in os.listdir(train_img_path):
    if filename.endswith('.jpg'):
        # Extract the texture features
        energy, contrast, correlation, entropy, mean, std_dev = extract_texture_features(os.path.join(train_img_path, filename))

        # Add the features to the dataframe
        df_train_img = pd.concat([df_train_img, pd.DataFrame({'PetID': [filename[:-4]], 'Energy': [energy], 'Contrast': [contrast], 'Correlation': [correlation], 'Entropy': [entropy], 'Mean': [mean], 'Std_Dev': [std_dev]})])

#df_train_img.to_csv('df_train_img_texture.csv')
# saving the files
df_train_img.to_csv(os.path.join(output_dir, 'df_train_img_texture.csv'), index=False)


# looping through the testing images
# Create an empty dataframe
df_test_img = pd.DataFrame(columns=['PetID', 'Energy', 'Contrast', 'Correlation', 'Entropy', 'Mean', 'Std_Dev'])

# Loop through each image in the folder
for filename in os.listdir(test_img_path):
    if filename.endswith('.jpg'):
        # Extract the texture features
        energy, contrast, correlation, entropy, mean, std_dev = extract_texture_features(os.path.join(test_img_path, filename))

        # Add the features to the dataframe
        df_test_img = pd.concat([df_test_img, pd.DataFrame({'PetID': [filename[:-4]], 'Energy': [energy], 'Contrast': [contrast], 'Correlation': [correlation], 'Entropy': [entropy], 'Mean': [mean], 'Std_Dev': [std_dev]})])

df_test_img.to_csv(os.path.join(output_dir, 'df_test_img_texture.csv'), index=False)

C:\Users\leoko\AppData\Local\Temp\ipykernel_7904\2979515395.py:34: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.
  df_train_img = pd.concat([df_train_img, pd.DataFrame({'PetID': [filename[:-4]], 'Energy': [energy], 'Contrast': [contrast], 'Correlation': [correlation], 'Entropy': [entropy], 'Mean': [mean], 'Std_Dev': [std_dev]})])
C:\Users\leoko\AppData\Local\Temp\ipykernel_7904\2979515395.py:52: FutureWarning: The behavior of DataFrame concatenation with empty or all-NA entries is deprecated. In a future version, this will no longer exclude empty or all-NA columns when determining the result dtypes. To retain the old behavior, exclude the relevant entries before the concat operation.
  df_test_img = pd.concat([df_test_img, pd.DataFrame({'PetID': [filename[:-4]], 'Energy': [energy], 'Contrast': [contrast], 'Correlation': [correlation], 'Entropy': [entropy], 'Mean': [mean], 'Std_Dev': [std_dev]})])

Extracting Local Features

def extract_keypoints(file_path):
    # Load an image
    image = cv2.imread(file_path, cv2.IMREAD_GRAYSCALE)

    # Initialize the SIFT detector
    sift = cv2.SIFT_create()

    # Detect and compute SIFT keypoints and descriptors
    keypoints, _ = sift.detectAndCompute(image, None)

    # Return the number of keypoints
    return len(keypoints)


df_train_img['num_checkpoints'] = None  # Initialize the column

def extract_and_update(row):
    file_path = os.path.join(train_img_path, f"{row['PetID']}.jpg")
    num_keypoints = extract_keypoints(file_path)
    return num_keypoints


# Update the 'num_checkpoints' column
df_train_img['num_checkpoints'] = df_train_img.apply(extract_and_update, axis=1)
df_train_img.to_csv('train_img_keypoints.csv')



# Assuming df is your DataFrame
df_test_img['num_checkpoints'] = None  # Initialize the column

def extract_and_update(row):
    file_path = os.path.join(test_img_path, f"{row['PetID']}.jpg")
    num_keypoints = extract_keypoints(file_path)
    return num_keypoints


# Update the 'num_checkpoints' column
df_test_img['num_checkpoints'] = df_test_img.apply(extract_and_update, axis=1)
df_test_img.to_csv(os.path.join(output_dir, 'test_img_keypoints.csv'), index=False)

df_train_img.to_csv(os.path.join(output_dir, 'train_img_keypoints.csv'), index=False)

Combining the current image feature dataframes first

train_texture = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\df_train_img_texture.csv')
test_texture = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\df_test_img_texture.csv')
train_keypoints =  pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\train_img_keypoints.csv')
test_keypoints = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\test_img_keypoints.csv')

test_texture.shape

(14465, 7)

train_texture.columns

Index(['PetID', 'Energy', 'Contrast', 'Correlation', 'Entropy', 'Mean',
       'Std_Dev'],
      dtype='object')

train_keypoints.columns

Index(['PetID', 'Energy', 'Contrast', 'Correlation', 'Entropy', 'Mean',
       'Std_Dev', 'num_checkpoints'],
      dtype='object')

final_keypoints = pd.concat([train_keypoints, test_keypoints], ignore_index=True, sort=False)
final_keypoints.head()

	PetID	Energy	Contrast	Correlation	Entropy	Mean	Std_Dev	num_checkpoints
0	0008c5398-1	0.024081	78.820303	0.989859	7.601080	143.223922	70.638848	319
1	0008c5398-2	0.070913	41.355089	0.994805	7.284530	91.658592	64.059420	233
2	0008c5398-3	0.040957	66.469614	0.989986	7.594550	111.068160	73.856578	410
3	0008c5398-4	0.050778	150.613044	0.984973	7.517500	108.980235	71.533391	469
4	0008c5398-5	0.025575	79.317245	0.992683	7.719211	141.169160	76.906123	395

# Columns to drop
columns_to_drop = ['mean_h', 'mean_s', 'mean_v', 'variance_h', 'variance_s', 'variance_v', 'Unnamed: 0']

# Drop specified columns
final_keypoints = final_keypoints.drop(columns=columns_to_drop, errors='ignore')

# Print the resulting DataFrame
final_keypoints.head()

	PetID	Energy	Contrast	Correlation	Entropy	Mean	Std_Dev	num_checkpoints
0	0008c5398-1	0.024081	78.820303	0.989859	7.601080	143.223922	70.638848	319
1	0008c5398-2	0.070913	41.355089	0.994805	7.284530	91.658592	64.059420	233
2	0008c5398-3	0.040957	66.469614	0.989986	7.594550	111.068160	73.856578	410
3	0008c5398-4	0.050778	150.613044	0.984973	7.517500	108.980235	71.533391	469
4	0008c5398-5	0.025575	79.317245	0.992683	7.719211	141.169160	76.906123	395

#final_keypoints.to_csv('image_features.csv')
final_keypoints.to_csv(os.path.join(output_dir, 'image_features.csv'), index=False)

• Generating Preliminary Features CSV

df_image = pd.read_csv("C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\image_features.csv", header = 0)

df_train.columns

Index(['Type', 'Name', 'Age', 'Breed1', 'Breed2', 'Gender', 'Color1', 'Color2',
       'Color3', 'MaturitySize', 'FurLength', 'Vaccinated', 'Dewormed',
       'Sterilized', 'Health', 'Quantity', 'Fee', 'State', 'RescuerID',
       'VideoAmt', 'Description', 'PetID', 'PhotoAmt', 'AdoptionSpeed',
       'GenderName', 'Breed1AgeFurLength', 'VDSCombination', 'ColorCount',
       'TotalVisualMedia', 'DescriptionLength', 'SentimentScore'],
      dtype='object')

# Adding features engineering for train dataset

# Concatenating Breed1, Age and Furlength features
df_train['Breed1AgeFurLength'] = (df_train['Breed1'].astype(str) + np.abs(df_train['Age']).astype(str) + df_train['FurLength'].astype(str)).astype(float)

# Concatenating binary features such as Vaccinated, Dewormed and Sterilized
df_train['VDSCombination'] = (df_train['Vaccinated'].astype(str) + df_train['Dewormed'].astype(str) + df_train['Sterilized'].astype(str)).astype(float)

# Creating a color count feature without including cases where any color is 0
df_train['ColorCount'] = df_train[['Color1', 'Color2', 'Color3']].apply(lambda row: len([color for color in row if color != 0]), axis=1)

# Creating a total visual media feature
df_train['TotalVisualMedia'] = df_train['PhotoAmt'] + df_train['VideoAmt']

# Creating a description length feature
df_train['DescriptionLength'] = df_train['Description'].apply(lambda x: len(str(x)))

# Creating a sentiment score feature
from textblob import TextBlob
df_train['SentimentScore'] = df_train['Description'].apply(lambda x: TextBlob(str(x)).sentiment.polarity)

# Adding features engineering for test dataset

# Concatenating Breed1, Age and Furlength features
df_test['Breed1AgeFurLength'] = (df_test['Breed1'].astype(str) + np.abs(df_test['Age']).astype(str) + df_test['FurLength'].astype(str)).astype(float)

# Concatenating binary features such as Vaccinated, Dewormed and Sterilized
df_test['VDSCombination'] = (df_test['Vaccinated'].astype(str) + df_test['Dewormed'].astype(str) + df_test['Sterilized'].astype(str)).astype(float)

# Creating a color count feature without including cases where any color is 0
df_test['ColorCount'] = df_test[['Color1', 'Color2', 'Color3']].apply(lambda row: len([color for color in row if color != 0]), axis=1)

# Creating a total visual media feature
df_test['TotalVisualMedia'] = df_test['PhotoAmt'] + df_test['VideoAmt']

# Creating a description length feature
df_test['DescriptionLength'] = df_test['Description'].apply(lambda x: len(str(x)))

# Creating a sentiment score feature
from textblob import TextBlob
df_test['SentimentScore'] = df_test['Description'].apply(lambda x: TextBlob(str(x)).sentiment.polarity)

# Adding image features to dataset

# First finding the position of '-' in PetID
df_image['pos'] = df_image['PetID'].str.find('-')
#df_image.drop(columns=['Unnamed: 0'], axis=1, inplace=True)

# Using position to slice PetID using a lambda function
df_image['PetID'] = df_image.apply(lambda x: x['PetID'][0:x['pos']],axis=1)
df_image.drop(columns=['pos'], axis=1, inplace=True)

df_train = pd.merge(df_train, df_image, how="inner", on=["PetID"], copy=True)

df_train.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 58311 entries, 0 to 58310
Data columns (total 38 columns):
 #   Column              Non-Null Count  Dtype  
---  ------              --------------  -----  
 0   Type                58311 non-null  int64  
 1   Name                58311 non-null  object 
 2   Age                 58311 non-null  float64
 3   Breed1              58311 non-null  int64  
 4   Breed2              58311 non-null  int64  
 5   Gender              58311 non-null  int64  
 6   Color1              58311 non-null  int64  
 7   Color2              58311 non-null  int64  
 8   Color3              58311 non-null  int64  
 9   MaturitySize        58311 non-null  int64  
 10  FurLength           58311 non-null  int64  
 11  Vaccinated          58311 non-null  int64  
 12  Dewormed            58311 non-null  int64  
 13  Sterilized          58311 non-null  int64  
 14  Health              58311 non-null  int64  
 15  Quantity            58311 non-null  float64
 16  Fee                 58311 non-null  float64
 17  State               58311 non-null  int64  
 18  RescuerID           58311 non-null  object 
 19  VideoAmt            58311 non-null  int64  
 20  Description         58311 non-null  object 
 21  PetID               58311 non-null  object 
 22  PhotoAmt            58311 non-null  float64
 23  AdoptionSpeed       58311 non-null  int64  
 24  GenderName          58311 non-null  object 
 25  Breed1AgeFurLength  58311 non-null  float64
 26  VDSCombination      58311 non-null  float64
 27  ColorCount          58311 non-null  int64  
 28  TotalVisualMedia    58311 non-null  float64
 29  DescriptionLength   58311 non-null  int64  
 30  SentimentScore      58311 non-null  float64
 31  Energy              58311 non-null  float64
 32  Contrast            58311 non-null  float64
 33  Correlation         58311 non-null  float64
 34  Entropy             58311 non-null  float64
 35  Mean                58311 non-null  float64
 36  Std_Dev             58311 non-null  float64
 37  num_checkpoints     58311 non-null  int64  
dtypes: float64(14), int64(19), object(5)
memory usage: 16.9+ MB

numeric_cols = df_train.select_dtypes(include=[np.number])
correlation_m = numeric_cols.corr()

# Compute the correlation between features and the target variable (e.g., 'AdoptionSpeed')
corr_to_target = numeric_cols.corr()['AdoptionSpeed'].abs().sort_values(ascending=False)

# Display the top N features with the highest absolute correlation to the target
top_n = 16 # Change this number to display a different number of top features
top_features_to_target = corr_to_target[1:top_n]  # Exclude the target variable
print("Top", top_n, "features with highest correlation to the target:")
print(top_features_to_target)

Top 16 features with highest correlation to the target:
Sterilized            0.142998
Age                   0.114441
Vaccinated            0.112167
Type                  0.112119
VDSCombination        0.111070
FurLength             0.097280
Breed1                0.092313
num_checkpoints       0.085741
Contrast              0.085530
Correlation           0.085019
Breed1AgeFurLength    0.078735
Quantity              0.063614
Dewormed              0.062410
MaturitySize          0.057874
Gender                0.052376
Name: AdoptionSpeed, dtype: float64

# Columns to add

# Columns to drop (To be analyze)
# df_train.drop(columns=['Breed1', 'Dewormed', 'Quantity', 'Sterilized'], inplace=True)

# Dropping label GenderName
df_train.drop('GenderName', axis=1, inplace=True)

# Print the resulting DataFrame
df_train.head()

# Save the resulting in a csv file
df_train.to_csv(os.path.join(output_dir, 'preliminary_features.csv'), index=False)

• Making new features and data transformations

#Loading dataset for making new features

#Prelimary data
df_prey = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\preliminary_features.csv', header=0)

#Image data
df_imgs = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\image_features.csv', header=0)
df_imgs = df_imgs[['PetID', 'Energy', 'Contrast', 'Correlation', 'Entropy', 'Mean', 'Std_Dev', 'num_checkpoints']]

#Feature engineering on tabular data
df_tabr = pd.read_csv('C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\train_df_mahmood_features.csv')
df_tabr = df_tabr[['PetID', 'num_breed', 'num_colors', 'Total_Vaccinations', 'Avg_Health_Score', 'Fur_Health_Interact', 'Fur_Health_Interact_Encoded', 'Overall_Health_Status', 'Overall_Health_Status_Encoded']]

#Sentiment data
df_sent = pd.read_csv( 'C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\processed_train.csv', header=0)
df_sent = df_sent[['PetID', 'polarity', 'subjectivity']]

#Displaying first 5 rows
df_petf = pd.merge(df_prey, df_sent, how="inner", on=["PetID"], copy=True)
df_petf = pd.merge(df_petf, df_tabr, how="inner", on=["PetID"], copy=True)
df_petf.describe().T

	count	mean	std	min	25%	50%	75%	max
Type	58311.0	1.479669	0.499591	1.000000	1.000000	1.000000	2.000000	2.000000
Age	58311.0	-0.075961	0.911200	-0.523092	-0.467960	-0.412827	-0.137164	13.480591
Breed1	58311.0	267.609628	56.913768	0.000000	265.000000	266.000000	307.000000	307.000000
Breed2	58311.0	79.078784	125.182464	0.000000	0.000000	0.000000	218.000000	307.000000
Gender	58311.0	1.836326	0.713444	1.000000	1.000000	2.000000	2.000000	3.000000
Color1	58311.0	2.169505	1.701183	1.000000	1.000000	1.000000	2.000000	7.000000
Color2	58311.0	3.368215	2.687411	0.000000	0.000000	3.000000	6.000000	7.000000
Color3	58311.0	2.159404	3.099248	0.000000	0.000000	0.000000	6.000000	7.000000
MaturitySize	58311.0	1.869664	0.536435	1.000000	2.000000	2.000000	2.000000	4.000000
FurLength	58311.0	1.450532	0.588856	1.000000	1.000000	1.000000	2.000000	3.000000
Vaccinated	58311.0	1.699645	0.640694	1.000000	1.000000	2.000000	2.000000	3.000000
Dewormed	58311.0	1.496939	0.659212	1.000000	1.000000	1.000000	2.000000	3.000000
Sterilized	58311.0	1.882715	0.542593	1.000000	2.000000	2.000000	2.000000	3.000000
Health	58311.0	1.032327	0.189597	1.000000	1.000000	1.000000	1.000000	3.000000
Quantity	58311.0	0.123604	1.144796	-0.391224	-0.391224	-0.391224	0.287903	12.512201
Fee	58311.0	0.005178	1.009167	-0.517919	-0.243348	-0.243348	-0.243348	40.125697
State	58311.0	41346.082849	32.568108	41324.000000	41326.000000	41326.000000	41401.000000	41415.000000
VideoAmt	58311.0	0.127300	0.595185	0.000000	0.000000	0.000000	0.000000	8.000000
PhotoAmt	58311.0	7.016841	5.795156	1.000000	3.000000	5.000000	9.000000	30.000000
AdoptionSpeed	58311.0	2.491914	1.123897	0.000000	2.000000	2.000000	3.000000	4.000000
Breed1AgeFurLength	58311.0	2690.344570	826.376396	0.357695	2650.523092	2660.523092	3070.412827	30713.480591
VDSCombination	58311.0	186.816604	69.105185	111.000000	112.000000	212.000000	222.000000	333.000000
ColorCount	58311.0	2.077087	0.767663	1.000000	1.000000	2.000000	3.000000	3.000000
TotalVisualMedia	58311.0	7.144141	5.985989	1.000000	3.000000	5.000000	9.000000	38.000000
DescriptionLength	58311.0	434.146559	479.781642	1.000000	149.000000	304.000000	535.000000	6664.000000
SentimentScore	58311.0	0.199407	0.170476	-1.000000	0.081217	0.188235	0.301786	1.000000
Energy	58311.0	0.034879	0.040149	0.006036	0.018295	0.025374	0.037113	0.961818
Contrast	58311.0	199.383999	240.076671	1.313901	59.226359	116.042500	239.566469	3247.724868
Correlation	58311.0	0.968525	0.038533	0.465489	0.962978	0.981597	0.990214	0.999450
Entropy	58311.0	7.391165	0.405306	0.499963	7.228698	7.484712	7.663820	7.971661
Mean	58311.0	117.984377	27.888895	5.938269	100.767600	118.303059	135.119820	252.709617
Std_Dev	58311.0	59.752222	12.851703	5.488057	51.207725	59.862779	68.323232	115.744604
num_checkpoints	58311.0	668.592976	580.219031	4.000000	297.000000	499.000000	843.000000	8852.000000
polarity	58311.0	0.199458	0.170401	-1.000000	0.081217	0.188235	0.301786	1.000000
subjectivity	58311.0	0.505724	0.197572	0.000000	0.425505	0.522262	0.622500	1.000000
num_breed	58311.0	1.303528	0.459785	1.000000	1.000000	1.000000	2.000000	2.000000
num_colors	58311.0	2.077087	0.767663	1.000000	1.000000	2.000000	3.000000	3.000000
Total_Vaccinations	58311.0	5.079299	1.535524	3.000000	4.000000	5.000000	6.000000	9.000000
Avg_Health_Score	58311.0	1.032327	0.189597	1.000000	1.000000	1.000000	1.000000	3.000000
Fur_Health_Interact_Encoded	58311.0	1.383924	1.780581	0.000000	0.000000	0.000000	3.000000	8.000000
Overall_Health_Status_Encoded	58311.0	0.029994	0.170573	0.000000	0.000000	0.000000	0.000000	1.000000

• Univariate Analysis for Outliers Detection

#Collecting numerical features for Univariate Analysis
numeric_cols = ['Age', 'Fee', 'Quantity', 'DescriptionLength',
            'SentimentScore', 'Energy', 'Contrast', 'num_checkpoints', 'polarity', 'subjectivity']

#Univariate Analysis
for column in numeric_cols:
    print(column)
    print('Skew :', round(df_petf[column].skew(), 2))
    plt.figure(figsize = (15, 4))
    plt.subplot(1, 2, 1)
    df_petf[column].hist(grid=False)
    plt.ylabel('count')
    plt.subplot(1, 2, 2)
    sns.boxplot(x=df_petf[column])
    plt.show()

Age
Skew : 4.32

Fee
Skew : 10.58

Quantity
Skew : 3.68

DescriptionLength
Skew : 3.51

SentimentScore
Skew : 0.3

Energy
Skew : 7.18

Contrast
Skew : 3.13

num_checkpoints
Skew : 2.76

polarity
Skew : 0.3

subjectivity
Skew : -0.7

Data Transformation

#Loading libraries
from scipy.stats import boxcox

#Using Log transformation for numeric columns
df_petf['Age_log'] = np.log1p(df_petf['Age'])
df_petf['Fee_log'] = np.log1p(df_petf['Fee'])
df_petf['Quantity_log'] = np.log1p(df_petf['Quantity'])
df_petf['DescriptionLength_log'] = np.log1p(df_petf['DescriptionLength'])
df_petf['Energy_log'] = np.log1p(df_petf['Energy'])
df_petf['Contrast_log'] = np.log1p(df_petf['Contrast'])
df_petf['num_checkpoints_log'] = np.log1p(df_petf['num_checkpoints'])

#Using Box-Cox transformation for numeric columns
df_petf['Age_boxcox'], lambda_value = boxcox(df_petf['Age'] + 1)
df_petf['Quantity_boxcox'], lambda_value = boxcox(df_petf['Quantity'] + 1)
df_petf['Fee_boxcox'], lambda_value = boxcox(df_petf['Fee'] + 1)
df_petf['Energy_boxcox'], lambda_value = boxcox(df_petf['Energy'] + 1)

#Using Exponential transformation for emphasizing the differences in small values
df_petf['SentimentScore_exp'] = np.exp(df_petf['SentimentScore'])
df_petf['polarity_exp'] = np.exp(df_petf['polarity'])
df_petf['subjectivity_exp'] = np.exp(df_petf['subjectivity'])
df_petf['Energy_exp'], lambda_value = boxcox(df_petf['Energy'] + 1)

#Using Square root transformation to mitigate the impact of large values
df_petf['DescriptionLength_sqrt'] = np.sqrt(df_petf['DescriptionLength'])
df_petf['Contrast_sqrt'] = np.sqrt(df_petf['Contrast'])
df_petf['num_checkpoints_sqrt'] = np.sqrt(df_petf['num_checkpoints'])

Text Data Processing (NLP)

import nltk
nltk.download('punkt')
nltk.download('stopwords')

[nltk_data] Downloading package punkt to
[nltk_data]     C:\Users\leoko\AppData\Roaming\nltk_data...
[nltk_data]   Package punkt is already up-to-date!
[nltk_data] Downloading package stopwords to
[nltk_data]     C:\Users\leoko\AppData\Roaming\nltk_data...
[nltk_data]   Package stopwords is already up-to-date!

True

from nltk.tokenize import word_tokenize
from nltk.corpus import stopwords

#Specify the target word for popularity (lowercased)
target_word = "language"

#Tokenization function
def tokenize(text):
    tokens = word_tokenize(text.lower())
    tokens = [token for token in tokens if token.isalpha()]
    stop_words = set(stopwords.words('english'))
    tokens = [token for token in tokens if token not in stop_words]
    return tokens

#Tokenize the text
df_petf['Tokenized_Text'] = df_petf['Description'].apply(tokenize)

# Calculate popularity for the target word
df_petf['Popularity'] = df_petf['Tokenized_Text'].apply(lambda tokens: tokens.count(target_word))

# Checking for new features created by data transformation
df_petf.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 58311 entries, 0 to 58310
Data columns (total 67 columns):
 #   Column                         Non-Null Count  Dtype  
---  ------                         --------------  -----  
 0   Type                           58311 non-null  int64  
 1   Name                           58311 non-null  object 
 2   Age                            58311 non-null  float64
 3   Breed1                         58311 non-null  int64  
 4   Breed2                         58311 non-null  int64  
 5   Gender                         58311 non-null  int64  
 6   Color1                         58311 non-null  int64  
 7   Color2                         58311 non-null  int64  
 8   Color3                         58311 non-null  int64  
 9   MaturitySize                   58311 non-null  int64  
 10  FurLength                      58311 non-null  int64  
 11  Vaccinated                     58311 non-null  int64  
 12  Dewormed                       58311 non-null  int64  
 13  Sterilized                     58311 non-null  int64  
 14  Health                         58311 non-null  int64  
 15  Quantity                       58311 non-null  float64
 16  Fee                            58311 non-null  float64
 17  State                          58311 non-null  int64  
 18  RescuerID                      58311 non-null  object 
 19  VideoAmt                       58311 non-null  int64  
 20  Description                    58311 non-null  object 
 21  PetID                          58311 non-null  object 
 22  PhotoAmt                       58311 non-null  float64
 23  AdoptionSpeed                  58311 non-null  int64  
 24  Breed1AgeFurLength             58311 non-null  float64
 25  VDSCombination                 58311 non-null  float64
 26  ColorCount                     58311 non-null  int64  
 27  TotalVisualMedia               58311 non-null  float64
 28  DescriptionLength              58311 non-null  int64  
 29  SentimentScore                 58311 non-null  float64
 30  Energy                         58311 non-null  float64
 31  Contrast                       58311 non-null  float64
 32  Correlation                    58311 non-null  float64
 33  Entropy                        58311 non-null  float64
 34  Mean                           58311 non-null  float64
 35  Std_Dev                        58311 non-null  float64
 36  num_checkpoints                58311 non-null  int64  
 37  polarity                       58311 non-null  float64
 38  subjectivity                   58311 non-null  float64
 39  num_breed                      58311 non-null  int64  
 40  num_colors                     58311 non-null  int64  
 41  Total_Vaccinations             58311 non-null  int64  
 42  Avg_Health_Score               58311 non-null  float64
 43  Fur_Health_Interact            58311 non-null  object 
 44  Fur_Health_Interact_Encoded    58311 non-null  int64  
 45  Overall_Health_Status          58311 non-null  object 
 46  Overall_Health_Status_Encoded  58311 non-null  int64  
 47  Age_log                        58311 non-null  float64
 48  Fee_log                        58311 non-null  float64
 49  Quantity_log                   58311 non-null  float64
 50  DescriptionLength_log          58311 non-null  float64
 51  Energy_log                     58311 non-null  float64
 52  Contrast_log                   58311 non-null  float64
 53  num_checkpoints_log            58311 non-null  float64
 54  Age_boxcox                     58311 non-null  float64
 55  Quantity_boxcox                58311 non-null  float64
 56  Fee_boxcox                     58311 non-null  float64
 57  Energy_boxcox                  58311 non-null  float64
 58  SentimentScore_exp             58311 non-null  float64
 59  polarity_exp                   58311 non-null  float64
 60  subjectivity_exp               58311 non-null  float64
 61  Energy_exp                     58311 non-null  float64
 62  DescriptionLength_sqrt         58311 non-null  float64
 63  Contrast_sqrt                  58311 non-null  float64
 64  num_checkpoints_sqrt           58311 non-null  float64
 65  Tokenized_Text                 58311 non-null  object 
 66  Popularity                     58311 non-null  int64  
dtypes: float64(35), int64(25), object(7)
memory usage: 29.8+ MB

#Printing boxcox transformation for feature 'Age_log'
sns.displot(data = df_petf, x = "Age_boxcox", kde=True)

#Printing exponential transformation for feature 'Energy_log'
sns.displot(data = df_petf, x = "Energy_exp", kde=True)

#Printing popularity
sns.displot(data = df_petf, x = "Popularity", kde=True)

<seaborn.axisgrid.FacetGrid at 0x1e625846210>

Bivariate Analysis

# Let's consider these features to identify correlations between each others in plots
features = ['Age_log', 'Fee_boxcox', 'DescriptionLength_sqrt',
            'Energy_exp', 'Contrast_sqrt', 'num_checkpoints_sqrt', 'Popularity']
plt.figure(figsize=(13,17))
sns.pairplot(data=df_petf[features])
plt.show()

<Figure size 1300x1700 with 0 Axes>

Multivariate Analysis

# detecting multi co linearity to stop data leakage

# Compute the correlation matrix between selected features
numeric_cols = ['Age_boxcox', 'Fee_boxcox', 'Quantity_boxcox', 'DescriptionLength_sqrt',
                'Energy_exp', 'Contrast_sqrt', 'num_checkpoints_sqrt',
                'SentimentScore_exp', 'polarity_exp', 'subjectivity_exp',
                'Vaccinated', 'Type', 'VDSCombination', 'FurLength', 'Gender',
                'Popularity',
                'AdoptionSpeed']

new_df_petf = pd.DataFrame(df_petf, columns=numeric_cols)
correlation_matrix = new_df_petf.corr(method='spearman').round(2)

plt.figure(figsize=(17,10))
sns.heatmap(correlation_matrix, cmap='coolwarm', linewidth = 0.5, annot=True)
plt.title('Correlation Heatmap')
plt.show()

#Checking for new numeric values
new_df_petf['Popularity'].isnull().sum()

0

# Import library for machine learning
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, confusion_matrix

# Separate the target variable 'AdoptionSpeed from the features
predictor_variables = ['Age_boxcox', 'Fee_boxcox', 'Quantity_boxcox', 'DescriptionLength_sqrt',
                'Energy_exp', 'Contrast_sqrt', 'num_checkpoints_sqrt',
                'SentimentScore_exp', 'polarity_exp', 'subjectivity_exp',
                'Vaccinated', 'Type', 'VDSCombination', 'FurLength', 'Gender']
target_variable = ['AdoptionSpeed']

x = new_df_petf[predictor_variables]
y = new_df_petf[target_variable]

# Split the dataset into training and testing sets (e.g., 80% train, 20% test)
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.30, random_state=42)

# Scale the features using StandardScaler
# from sklearn.preprocessing import StandardScaler
# scaler = StandardScaler()

# # Transforming "x_train"
# x_train = scaler.fit_transform(x_train)

# # Transforming "x_test"
# x_test = scaler.transform(x_test)

pd.DataFrame(x_train, columns = predictor_variables).head()

	Age_boxcox	Fee_boxcox	Quantity_boxcox	DescriptionLength_sqrt	Energy_exp	Contrast_sqrt	num_checkpoints_sqrt	SentimentScore_exp	polarity_exp	subjectivity_exp	Vaccinated	Type	VDSCombination	FurLength	Gender
7654	-0.886592	-0.42356	0.192552	44.519659	0.008522	32.924460	41.315857	1.261573	1.261573	1.730944	2	1	212.0	1	1
43634	0.353544	-0.42356	-0.917684	60.456596	0.017914	13.762494	22.956481	1.053163	1.053163	1.652848	2	1	211.0	2	1
7361	-1.164078	-0.42356	-0.917684	22.271057	0.019990	8.030423	33.166248	1.062279	1.062279	1.711722	2	1	222.0	1	1
13315	-1.164078	-0.42356	-0.917684	9.949874	0.018661	13.643717	17.944358	1.542390	1.542390	2.117000	1	1	112.0	2	2
18551	-1.164078	-0.42356	-0.917684	12.845233	0.023762	8.571816	14.317821	1.150993	1.150993	1.504309	1	1	122.0	1	2

Approach on Class Separability

image_features=pd.read_csv("C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\image_features.csv")

#image_features=image_features.drop(image_features.columns[0], axis=1)
image_features.head()

	PetID	Energy	Contrast	Correlation	Entropy	Mean	Std_Dev	num_checkpoints
0	0008c5398-1	0.024081	78.820303	0.989859	7.601080	143.223922	70.638848	319
1	0008c5398-2	0.070913	41.355089	0.994805	7.284530	91.658592	64.059420	233
2	0008c5398-3	0.040957	66.469614	0.989986	7.594550	111.068160	73.856578	410
3	0008c5398-4	0.050778	150.613044	0.984973	7.517500	108.980235	71.533391	469
4	0008c5398-5	0.025575	79.317245	0.992683	7.719211	141.169160	76.906123	395

#clean petID column
image_features["PetID"]=image_features["PetID"].str.split("-", expand=True)[0]

image_features.head(3)

	PetID	Energy	Contrast	Correlation	Entropy	Mean	Std_Dev	num_checkpoints
0	0008c5398	0.024081	78.820303	0.989859	7.60108	143.223922	70.638848	319
1	0008c5398	0.070913	41.355089	0.994805	7.28453	91.658592	64.059420	233
2	0008c5398	0.040957	66.469614	0.989986	7.59455	111.068160	73.856578	410

#Group by ID
image_features_transformed=image_features.groupby("PetID", as_index=False).mean()

sentiment_features=pd.read_csv("C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\processed_train.csv")
sentiment_features.head()

	Type	Name	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	FurLength	Vaccinated	Dewormed	Sterilized	Health	Quantity	Fee	State	RescuerID	VideoAmt	Description	PetID	PhotoAmt	AdoptionSpeed	polarity	subjectivity
0	2	Nibble	3	299	0	1	1	7	0	1	1	2	2	2	1	1	100	41326	8480853f516546f6cf33aa88cd76c379	0	Nibble is a 3+ month old ball of cuteness. He ...	86e1089a3	1.0	2	0.130357	0.407143
1	2	No Name Yet	1	265	0	1	1	2	0	2	2	3	3	3	1	1	0	41401	3082c7125d8fb66f7dd4bff4192c8b14	0	I just found it alone yesterday near my apartm...	6296e909a	2.0	0	0.100000	0.400000
2	1	Brisco	1	307	0	1	2	7	0	2	2	1	1	2	1	1	0	41326	fa90fa5b1ee11c86938398b60abc32cb	0	Their pregnant mother was dumped by her irresp...	3422e4906	7.0	3	0.261905	0.542857
3	1	Miko	4	307	0	2	1	2	0	2	1	1	1	2	1	1	150	41401	9238e4f44c71a75282e62f7136c6b240	0	Good guard dog, very alert, active, obedience ...	5842f1ff5	8.0	2	0.341146	0.437500
4	1	Hunter	1	307	0	1	1	0	0	2	1	2	2	2	1	1	0	41326	95481e953f8aed9ec3d16fc4509537e8	0	This handsome yet cute boy is up for adoption....	850a43f90	3.0	2	0.400000	0.743750

sentiment_features.shape

(14993, 26)

image_features.shape, image_features_transformed.shape

((72776, 8), (18510, 8))

all_data=sentiment_features.merge(image_features_transformed, how="left", on="PetID")
all_data.head(3)

	Type	Name	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	FurLength	Vaccinated	Dewormed	Sterilized	Health	Quantity	Fee	State	RescuerID	VideoAmt	Description	PetID	PhotoAmt	AdoptionSpeed	polarity	subjectivity	Energy	Contrast	Correlation	Entropy	Mean	Std_Dev	num_checkpoints
0	2	Nibble	3	299	0	1	1	7	0	1	1	2	2	2	1	1	100	41326	8480853f516546f6cf33aa88cd76c379	0	Nibble is a 3+ month old ball of cuteness. He ...	86e1089a3	1.0	2	0.130357	0.407143	0.035852	157.460689	0.986806	7.433718	95.712315	77.438182	482.000000
1	2	No Name Yet	1	265	0	1	1	2	0	2	2	3	3	3	1	1	0	41401	3082c7125d8fb66f7dd4bff4192c8b14	0	I just found it alone yesterday near my apartm...	6296e909a	2.0	0	0.100000	0.400000	0.022545	63.340863	0.993091	7.714046	102.676512	68.204360	187.500000
2	1	Brisco	1	307	0	1	2	7	0	2	2	1	1	2	1	1	0	41326	fa90fa5b1ee11c86938398b60abc32cb	0	Their pregnant mother was dumped by her irresp...	3422e4906	7.0	3	0.261905	0.542857	0.030182	78.109415	0.982029	7.185988	142.957730	48.593305	296.857143

all_data.columns , len(all_data.columns)

(Index(['Type', 'Name', 'Age', 'Breed1', 'Breed2', 'Gender', 'Color1', 'Color2',
        'Color3', 'MaturitySize', 'FurLength', 'Vaccinated', 'Dewormed',
        'Sterilized', 'Health', 'Quantity', 'Fee', 'State', 'RescuerID',
        'VideoAmt', 'Description', 'PetID', 'PhotoAmt', 'AdoptionSpeed',
        'polarity', 'subjectivity', 'Energy', 'Contrast', 'Correlation',
        'Entropy', 'Mean', 'Std_Dev', 'num_checkpoints'],
       dtype='object'),
 33)

all_data[["Type",'Breed1', 'Breed2', 'Gender', 'Color1', 'Color2',
       'Color3', 'MaturitySize', 'FurLength', 'Vaccinated', 'Dewormed',
       'Sterilized', 'Health',"State","PetID"]].head()

	Type	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	FurLength	Vaccinated	Dewormed	Sterilized	Health	State	PetID
0	2	299	0	1	1	7	0	1	1	2	2	2	1	41326	86e1089a3
1	2	265	0	1	1	2	0	2	2	3	3	3	1	41401	6296e909a
2	1	307	0	1	2	7	0	2	2	1	1	2	1	41326	3422e4906
3	1	307	0	2	1	2	0	2	1	1	1	2	1	41401	5842f1ff5
4	1	307	0	1	1	0	0	2	1	2	2	2	1	41326	850a43f90

all_data["FurLength"].unique()

array([1, 2, 3], dtype=int64)

import pandas as pd
from sklearn.preprocessing import OneHotEncoder
from sklearn.decomposition import PCA
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt
import os

# Create a dummy dataframe with both categorical and numerical columns

columns_to_drop=["Name","Description","RescuerID"]
categorical_columns=["Type",'Breed1', 'Breed2', 'Gender', 'Color1', 'Color2',
       'Color3', 'MaturitySize', 'FurLength', 'Vaccinated', 'Dewormed',
       'Sterilized', 'Health',"State","PetID"]
numerical_columns=['polarity', 'subjectivity', 'Energy', 'Contrast', 'Correlation',
       'Entropy', 'Mean', 'Std_Dev', 'num_checkpoints','PhotoAmt','VideoAmt',"Age","Quantity","Fee"]

target=all_data["AdoptionSpeed"]
all_data=all_data.drop("AdoptionSpeed", axis=1)



#simple filling
all_data[categorical_columns] = all_data[categorical_columns].fillna("unknown")
all_data.fillna(all_data[numerical_columns].median(), inplace=True)
#all_data=all_data.fillna(0)

#transformation
all_data[categorical_columns]=all_data[categorical_columns].astype(str)
all_data[numerical_columns]=all_data[numerical_columns].astype(float)

all_data=all_data.drop(columns_to_drop, axis=1)

#normalize numerical columns:

for column in numerical_columns:
  all_data[numerical_columns] = (all_data[numerical_columns] - all_data[numerical_columns].mean()) / all_data[numerical_columns].std()


# Perform one-hot encoding on the categorical columns
encoder = OneHotEncoder(sparse=False)
encoded_cat = encoder.fit_transform(all_data[categorical_columns])
encoded_cat = pd.DataFrame(encoded_cat, columns=encoder.get_feature_names_out(categorical_columns))

# Concatenate the encoded categorical columns and the numerical columns
X = pd.concat([encoded_cat, all_data[numerical_columns]], axis=1)

# Perform PCA on the combined data
pca = PCA(n_components=14, )
pca_data = pca.fit_transform(X)

C:\Users\leoko\AppData\Roaming\Python\Python312\site-packages\sklearn\preprocessing\_encoders.py:975: FutureWarning: `sparse` was renamed to `sparse_output` in version 1.2 and will be removed in 1.4. `sparse_output` is ignored unless you leave `sparse` to its default value.
  warnings.warn(

pca_data = pd.DataFrame(pca_data)

print(pca.explained_variance_ratio_)

print(pca.components_)

[0.11663179 0.07489694 0.06708838 0.06452002 0.0578784  0.05402035
 0.04491721 0.04183083 0.04082405 0.03427079 0.03340837 0.02477002
 0.02339944 0.02220482]
[[ 8.17089287e-02 -8.17089287e-02  7.04708644e-05 ...  2.05823728e-02
   3.70447588e-02 -9.81414965e-03]
 [ 1.13474081e-01 -1.13474081e-01 -3.74956977e-06 ...  3.73812920e-01
  -3.83703729e-01  1.78497326e-01]
 [ 3.06233806e-02 -3.06233806e-02  3.41109048e-05 ...  2.93767985e-01
  -2.44279370e-01  1.66007903e-01]
 ...
 [ 4.43938561e-02 -4.43938561e-02 -1.95558152e-05 ... -1.53030875e-01
   2.79198930e-02 -1.59547607e-03]
 [ 1.22528823e-01 -1.22528823e-01  1.86053526e-04 ...  2.66243782e-01
  -1.61964972e-01  7.03764545e-02]
 [ 1.27293706e-02 -1.27293706e-02 -5.83100634e-05 ... -9.55114901e-03
   2.41578200e-01  3.25889193e-02]]

print(pca.explained_variance_ratio_.round(2)[:10])

[0.12 0.07 0.07 0.06 0.06 0.05 0.04 0.04 0.04 0.03]

variance_exp_cumsum = pca.explained_variance_ratio_.cumsum().round(2)
fig, axes = plt.subplots(1,1,figsize=(16,7), dpi=100)
plt.plot(variance_exp_cumsum, color='firebrick')
plt.title('Screeplot of Variance Explained %', fontsize=22)
plt.xlabel('# of PCs', fontsize=16)
plt.show()

pca_data.head()

	0	1	2	3	4	5	6	7	8	9	10	11	12	13
0	-0.604106	-0.819422	0.238892	-0.270389	-1.306638	-0.652097	0.269963	1.937159	-0.959592	0.485091	0.412230	-0.798072	0.259926	0.609048
1	-0.714525	-0.902383	0.250052	-1.253288	-1.219482	-0.490955	-0.102255	0.202045	-0.795531	0.009971	0.168518	-0.713605	-0.124492	-0.655824
2	-1.092104	0.733364	-0.176084	-0.145397	0.439406	0.531051	-0.949138	-0.673335	1.607342	-0.937639	-0.503719	0.093268	0.176357	0.145848
3	-0.241502	0.377959	-0.447610	-0.798102	-0.194653	1.157786	0.500191	1.339629	0.282886	-1.667357	-0.187460	0.670160	-0.314100	-0.248092
4	-1.465296	1.043056	-1.362919	1.233223	-0.867129	-0.892260	-0.848456	-0.162946	0.943938	-0.136355	0.403343	-0.242270	0.779366	0.409931

from itertools import combinations
plt.figure(figsize=(15, 12))
plt.subplots_adjust(hspace=0.5)
plt.suptitle("PCA plots without feature engineering")
n=1
for col1, col2 in combinations(pca_data.columns[:4], 2):
    ax = plt.subplot(3, 2, n)

    ax.set_xlabel(col1)
    ax.set_ylabel(col2)
    ax.set_title("pca{} vs pca{}".format(col1, col2))
    plt.scatter(pca_data[col1], pca_data[col2],c=target, label=target.unique(),cmap='rainbow',
  alpha=0.7,edgecolors='b')
    #plt.show()
    n=n+1

#LDA
# necessary import
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import sklearn
from sklearn.preprocessing import StandardScaler, LabelEncoder
from sklearn.model_selection import train_test_split
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score, confusion_matrix
 

# divide the dataset into class and target variable
X = encoded_cat[:7000]
y = target[:7000]
 
# Preprocess the dataset and divide into train and test
le = LabelEncoder()
y = le.fit_transform(y)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
 
# apply Linear Discriminant Analysis
lda = LinearDiscriminantAnalysis(n_components=2)
X_train = lda.fit_transform(X_train, y_train)
X_test = lda.transform(X_test)
 
# plot the scatterplot
plt.scatter(
    X_train[:,0],X_train[:,1],c=y_train,cmap='rainbow',
  alpha=0.7,edgecolors='b'
)

<matplotlib.collections.PathCollection at 0x1e625c75c40>

import pandas as pd
all_data=pd.read_csv("C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\preliminary_features.csv")
#all_data.drop('Unnamed: 0', inplace=True, axis=1)

columns_to_drop=["Name","Description","RescuerID","PetID"]

import prince

all_data.set_index("AdoptionSpeed", inplace=True)
all_data.drop(columns_to_drop, axis=1, inplace=True)
famd = prince.FAMD(
    n_components=2,
    n_iter=30,
    copy=True,
    check_input=True,
    random_state=42,
    engine="sklearn",
    handle_unknown="error"
)
famd = famd.fit(all_data)

C:\Users\leoko\AppData\Roaming\Python\Python312\site-packages\prince\pca.py:175: PerformanceWarning: Concatenating sparse arrays with multiple fill values: '[-0.06086275832670119, -0.03286964624210119, -0.021518234567264212, -0.024847117039791002, -0.017569564951772036, -0.027779921372933013, -0.03513912990354407, -0.05958129358143515, -0.012423558519895501, -0.043036469134528424, -0.0869649096392685, -0.1964337078363004, -0.12790836355575197, -0.052708694855316096, -0.04479377726721073, -0.13834289277321493, -0.0669029101207173, -0.07027825980708814, -0.08784782475886017, -0.07857348313452016, -0.03928674156726008, -0.049694234079582005, -0.1454140242986409, -0.030431379163350595, -0.14327546627925056, -0.09296939901197068, -0.10901628357676502, -0.11386379463626935, -0.1652846959479808, -0.1842711517991869, -0.06573929248420238, -0.0372706755596865, -0.05122364402357136, -0.05555984274586603, -0.20639565532744622, -0.048116235248188546, -0.041204282171516456, -0.2747264280686591, -0.04648469950598534, -0.07244117209230616, -0.13609326491664522, -0.1867670373036489, -0.3485245508654304, -0.08051386131893848, -0.19955189807339635, -0.09542716369849215, -0.10759117283632103, -0.10541738971063219, -0.2591138567103497, -0.08607293826905685, -0.05415303610738823, -0.20147626176843145, -0.16006635450758072, -0.10169120916734624, -0.09044487124074649, -0.2932065093199302, -0.11181202667905951, -0.2518644782183013, -0.3189251124478906, -0.07136791021047098, -0.11111968549173205, -0.11654330934961299, -0.21518234567264205, -0.056931897318134674, -0.17917510906884293, -0.0978231976089037, -0.1273035924913871, -0.11042300347972872, -0.08240856434303291, -0.27133461372223944, -0.4307231792802846, -0.8906920367713848, -1.5369101846150255, -0.058271654674806495, -0.13495438912120508, -0.06804663245832261, -0.08872195399764747, -0.19286549281206736, -0.07556955625778757, -0.36369423833318937, -0.1332278226354609, -0.39599788282012865, -0.07758509808973277, -0.4625169202613171, -0.1629334424945662, -0.18882174200729357, -1.8800255488177384, -2.502907708644791, -0.09938846815916401, -0.08958755453442147, -0.09860893872630355, -0.1338058202414346, -0.20899660327892144, -0.18046261144271292, -0.12361284651454937, -0.0814667212472831, -0.22704881700481847, -0.11521134206997616, -0.149082702238746, -0.13665914371885052, -0.17169699474008918, -0.2704800143909067, -0.07349876339370877, -0.14055651961417628, -0.14273582042094196, -0.06917144638660747, -0.06455470370179263, -0.09213557589959345, -0.17744390799529494, -0.2800127882776288, -0.47111339553813536, -0.6478144242931552, -0.11851319493225806, -0.1948559014993591, -0.3340528325773108, -0.13438133180163223, -1.0284575623501333, -2.158984286690734, -1.4592790317425273, -0.7500578770691915, -0.6318939894405015, -0.7075977329163355, -0.6403859385806666, -0.5801948142945969, -1.5269862437929258, -1.4597020418338074, -0.6966061089300114, -0.7260079293679722, -0.863146789368363, -0.8356170795818147, -1.4428988502815605, -2.4442858071348676, -0.35467029355765545, -0.34451582843550044, -0.5529530371332525, -0.5223807208670416, -1.4840303369728391, -1.9086241427342685, -1.7393869302254314, -0.09623247049637709, -0.10830607221477646, -0.31916699728542886, -0.17960530202677488, -0.173929819278537, -0.12045086878866088, -0.11318400471212023, -0.11720361666969523, -0.11786022470178023, -0.10244728804714272, -0.0970310939036212, -0.10468276480446069, -0.11980845957463075, -0.12669593464241813, -0.09461500285494659, -0.1191625871628703, -0.12172551665340238, -0.1416503611666587, -0.10092946653787334, -0.12108986992412069, -0.08426067344659863, -0.11453955009195639, -0.12910940740358526, -0.109721897956547, -0.10319782766372311, -0.09129413749005177, -0.1411045003263453, -0.14434870574456563, -0.1125001071838422, -0.11587923950329639, -0.10687149136227977, -0.14752158380357364, -0.12485521789666652, -0.10614693052415868, -0.13722268897039844, -0.15812608456594832, -0.1516488580184267, -0.15763728560800633, -0.13264730648561454, -0.13029938101426072, -0.12608534827099704, -0.13889960686466507, -0.15113911251557513, -0.12298695466240424, -0.14647159477291244, -0.14699752678741754, -0.1590991773478165, -0.15215689581675296, -0.10016193093846795, -0.13206423857953103, -0.1355250233328698, -0.13089031094392264, -0.13147858496840475, -0.14219412654868036, -0.16667952823759807, -0.12547179062371489, -0.10394294800903411, -0.12851028859125516, -0.14488234418461232, -0.1531679162134967, -0.1297057588790961, -0.09379580992210837, -0.1400063941388144, -0.1377839293044148, -0.07954958867390825, -0.1459437674894786, -0.12423558519895503, -0.1624591083221647, -0.12235786125474328, -0.08333976411879904, -0.15861337719894467, -0.062117792599477514, -0.15665511227299805, -0.14381308717858027, -0.16054775816449063, -0.17259359088730017, -0.074541351119373, -0.07658395810674835, -0.08517162612027318, -0.06334796732120906, -2.314752675905066, -1.6767202052533208, -0.911333859069002, -2.3219760493114845, -1.7782187847990278, -0.6681056651546216, -1.774134615755013, -2.0440057636075175, -1.2940196302680502, -2.9546659597636546, -0.4989570848855575, -1.3934891488726886, -2.5110651866612894, -0.8583051674415462, -0.2396169191492615, -0.23735175225496621, -2.328878806053677, -0.6378502248957812, -0.48674354404173303, -0.37907125275171766, -0.2300875236947083, -0.5123870758632121, -1.5262279700747101, -1.3836520434384194, -2.4953112382314933, -0.9267841426374844, -2.1640184305743357, -2.07774498727216, -1.898610795047973, -2.1198464650405606, -0.9494882903871659, -2.8922620579227147, -0.6325043382637933, -0.33359047549402593, -0.1778782878570062]'. Picking the first and converting the rest.
  X = self.scaler_.transform(X.to_numpy())

famd.row_coordinates(all_data).head()

C:\Users\leoko\AppData\Roaming\Python\Python312\site-packages\prince\pca.py:175: PerformanceWarning: Concatenating sparse arrays with multiple fill values: '[-0.06086275832670119, -0.03286964624210119, -0.021518234567264212, -0.024847117039791002, -0.017569564951772036, -0.027779921372933013, -0.03513912990354407, -0.05958129358143515, -0.012423558519895501, -0.043036469134528424, -0.0869649096392685, -0.1964337078363004, -0.12790836355575197, -0.052708694855316096, -0.04479377726721073, -0.13834289277321493, -0.0669029101207173, -0.07027825980708814, -0.08784782475886017, -0.07857348313452016, -0.03928674156726008, -0.049694234079582005, -0.1454140242986409, -0.030431379163350595, -0.14327546627925056, -0.09296939901197068, -0.10901628357676502, -0.11386379463626935, -0.1652846959479808, -0.1842711517991869, -0.06573929248420238, -0.0372706755596865, -0.05122364402357136, -0.05555984274586603, -0.20639565532744622, -0.048116235248188546, -0.041204282171516456, -0.2747264280686591, -0.04648469950598534, -0.07244117209230616, -0.13609326491664522, -0.1867670373036489, -0.3485245508654304, -0.08051386131893848, -0.19955189807339635, -0.09542716369849215, -0.10759117283632103, -0.10541738971063219, -0.2591138567103497, -0.08607293826905685, -0.05415303610738823, -0.20147626176843145, -0.16006635450758072, -0.10169120916734624, -0.09044487124074649, -0.2932065093199302, -0.11181202667905951, -0.2518644782183013, -0.3189251124478906, -0.07136791021047098, -0.11111968549173205, -0.11654330934961299, -0.21518234567264205, -0.056931897318134674, -0.17917510906884293, -0.0978231976089037, -0.1273035924913871, -0.11042300347972872, -0.08240856434303291, -0.27133461372223944, -0.4307231792802846, -0.8906920367713848, -1.5369101846150255, -0.058271654674806495, -0.13495438912120508, -0.06804663245832261, -0.08872195399764747, -0.19286549281206736, -0.07556955625778757, -0.36369423833318937, -0.1332278226354609, -0.39599788282012865, -0.07758509808973277, -0.4625169202613171, -0.1629334424945662, -0.18882174200729357, -1.8800255488177384, -2.502907708644791, -0.09938846815916401, -0.08958755453442147, -0.09860893872630355, -0.1338058202414346, -0.20899660327892144, -0.18046261144271292, -0.12361284651454937, -0.0814667212472831, -0.22704881700481847, -0.11521134206997616, -0.149082702238746, -0.13665914371885052, -0.17169699474008918, -0.2704800143909067, -0.07349876339370877, -0.14055651961417628, -0.14273582042094196, -0.06917144638660747, -0.06455470370179263, -0.09213557589959345, -0.17744390799529494, -0.2800127882776288, -0.47111339553813536, -0.6478144242931552, -0.11851319493225806, -0.1948559014993591, -0.3340528325773108, -0.13438133180163223, -1.0284575623501333, -2.158984286690734, -1.4592790317425273, -0.7500578770691915, -0.6318939894405015, -0.7075977329163355, -0.6403859385806666, -0.5801948142945969, -1.5269862437929258, -1.4597020418338074, -0.6966061089300114, -0.7260079293679722, -0.863146789368363, -0.8356170795818147, -1.4428988502815605, -2.4442858071348676, -0.35467029355765545, -0.34451582843550044, -0.5529530371332525, -0.5223807208670416, -1.4840303369728391, -1.9086241427342685, -1.7393869302254314, -0.09623247049637709, -0.10830607221477646, -0.31916699728542886, -0.17960530202677488, -0.173929819278537, -0.12045086878866088, -0.11318400471212023, -0.11720361666969523, -0.11786022470178023, -0.10244728804714272, -0.0970310939036212, -0.10468276480446069, -0.11980845957463075, -0.12669593464241813, -0.09461500285494659, -0.1191625871628703, -0.12172551665340238, -0.1416503611666587, -0.10092946653787334, -0.12108986992412069, -0.08426067344659863, -0.11453955009195639, -0.12910940740358526, -0.109721897956547, -0.10319782766372311, -0.09129413749005177, -0.1411045003263453, -0.14434870574456563, -0.1125001071838422, -0.11587923950329639, -0.10687149136227977, -0.14752158380357364, -0.12485521789666652, -0.10614693052415868, -0.13722268897039844, -0.15812608456594832, -0.1516488580184267, -0.15763728560800633, -0.13264730648561454, -0.13029938101426072, -0.12608534827099704, -0.13889960686466507, -0.15113911251557513, -0.12298695466240424, -0.14647159477291244, -0.14699752678741754, -0.1590991773478165, -0.15215689581675296, -0.10016193093846795, -0.13206423857953103, -0.1355250233328698, -0.13089031094392264, -0.13147858496840475, -0.14219412654868036, -0.16667952823759807, -0.12547179062371489, -0.10394294800903411, -0.12851028859125516, -0.14488234418461232, -0.1531679162134967, -0.1297057588790961, -0.09379580992210837, -0.1400063941388144, -0.1377839293044148, -0.07954958867390825, -0.1459437674894786, -0.12423558519895503, -0.1624591083221647, -0.12235786125474328, -0.08333976411879904, -0.15861337719894467, -0.062117792599477514, -0.15665511227299805, -0.14381308717858027, -0.16054775816449063, -0.17259359088730017, -0.074541351119373, -0.07658395810674835, -0.08517162612027318, -0.06334796732120906, -2.314752675905066, -1.6767202052533208, -0.911333859069002, -2.3219760493114845, -1.7782187847990278, -0.6681056651546216, -1.774134615755013, -2.0440057636075175, -1.2940196302680502, -2.9546659597636546, -0.4989570848855575, -1.3934891488726886, -2.5110651866612894, -0.8583051674415462, -0.2396169191492615, -0.23735175225496621, -2.328878806053677, -0.6378502248957812, -0.48674354404173303, -0.37907125275171766, -0.2300875236947083, -0.5123870758632121, -1.5262279700747101, -1.3836520434384194, -2.4953112382314933, -0.9267841426374844, -2.1640184305743357, -2.07774498727216, -1.898610795047973, -2.1198464650405606, -0.9494882903871659, -2.8922620579227147, -0.6325043382637933, -0.33359047549402593, -0.1778782878570062]'. Picking the first and converting the rest.
  X = self.scaler_.transform(X.to_numpy())

component	0	1
AdoptionSpeed
2	4.214693	-5.816055
0	0.228647	-0.485286
0	0.459619	-1.291746
3	-4.942247	0.473450
3	-4.827923	0.736891

famd.plot(
    all_data[:1000],
    x_component=0,
    y_component=1
)

C:\Users\leoko\AppData\Roaming\Python\Python312\site-packages\prince\pca.py:175: PerformanceWarning: Concatenating sparse arrays with multiple fill values: '[nan, nan, nan, nan, nan, -0.1341640786499874, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.16431676725154984, -0.23237900077244503, nan, nan, nan, nan, -0.28460498941515416, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.36742346141747667, nan, nan, nan, nan, -0.34205262752974136, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.21213203435596426, nan, nan, nan, nan, nan, nan, -0.2683281572999748, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.46475800154489005, -0.09486832980505137, nan, -0.4347413023856832, nan, nan, nan, -0.18973665961010275, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.25099800796022265, nan, nan, nan, nan, nan, -0.41352146256270667, -0.99498743710662, -1.4293355099485916, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.4449719092257397, nan, nan, nan, nan, -0.31464265445104544, nan, nan, nan, nan, nan, nan, -0.3286335345030997, nan, nan, nan, nan, nan, -1.9877122528172935, -2.49559612117025, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.3, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.40249223594996214, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.5449770637375485, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.5282045058497703, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.9439279633531364, -2.157081361469706, -0.6841052550594827, -0.8432081593533117, -0.7707139547199078, -0.5366563145999496, -0.4929503017546495, -1.6127616066858732, -1.3281566172707193, -0.709929573971954, -0.697137002317335, -0.758946638440411, -0.7162401831787993, -1.5987495113369072, -2.5667099563448925, -1.4103191128251789, -1.9967473550752484, -1.5530614926653743, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.4837354648979129, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -0.3794733192202055, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, -2.379285607067802, -1.5674820573135757, -0.9391485505499118, -2.2965191050805562, -1.7467111953611565, -0.8215838362577492, -1.7621010186706096, -2.1297887219158618, -1.1657615536635269, -2.9728773940409985, nan, -1.2621410380777578, -2.557928849675065, -0.9295160030897801, nan, -0.354964786985977, -2.373604853382298, -0.5770615218501403, nan, nan, nan, -1.520855022676389, nan, -1.3974977638622539, -2.47204368893432, -0.9674709297958258, -2.2827614855696154, -1.946535383701, -2.054507240191672, -1.9763602910400726, -0.9343446901438462, -2.903962809679215, -0.5531726674375732, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan]'. Picking the first and converting the rest.
  X = self.scaler_.transform(X.to_numpy())

famd.column_contributions_.style.format('{:.2%}')

component	0	1
variable
Age	0.17%	0.00%
Quantity	0.34%	0.26%
Fee	0.03%	0.02%
PhotoAmt	0.03%	0.03%
Breed1AgeFurLength	0.01%	0.00%
VDSCombination	0.17%	0.00%
TotalVisualMedia	0.03%	0.03%
SentimentScore	0.02%	0.00%
Energy	0.00%	0.01%
Contrast	0.03%	0.11%
Correlation	0.03%	0.09%
Entropy	0.00%	0.07%
Mean	0.01%	0.01%
Std_Dev	0.00%	0.03%
Breed1	1.18%	1.76%
Breed2	0.59%	1.02%
Color1	0.53%	1.10%
Color2	1.37%	1.47%
Color3	1.44%	0.83%
ColorCount	1.66%	0.70%
DescriptionLength	0.83%	1.26%
Dewormed	0.48%	0.07%
FurLength	0.00%	0.09%
Gender	0.56%	0.25%
Health	0.00%	0.01%
MaturitySize	0.12%	0.18%
State	0.03%	0.12%
Sterilized	0.26%	0.04%
Type	1.03%	1.47%
Vaccinated	0.63%	0.07%
VideoAmt	0.01%	0.05%
num_checkpoints	0.28%	0.55%

Class Separability with feature engineering

import pandas as pd
all_data=pd.read_csv("C:\\Users\\leoko\\Downloads\\petfinder-adoption-prediction\\output\\preliminary_features.csv")
#all_data.drop('Unnamed: 0', inplace=True, axis=1)

all_data.columns

Index(['Type', 'Name', 'Age', 'Breed1', 'Breed2', 'Gender', 'Color1', 'Color2',
       'Color3', 'MaturitySize', 'FurLength', 'Vaccinated', 'Dewormed',
       'Sterilized', 'Health', 'Quantity', 'Fee', 'State', 'RescuerID',
       'VideoAmt', 'Description', 'PetID', 'PhotoAmt', 'AdoptionSpeed',
       'Breed1AgeFurLength', 'VDSCombination', 'ColorCount',
       'TotalVisualMedia', 'DescriptionLength', 'SentimentScore', 'Energy',
       'Contrast', 'Correlation', 'Entropy', 'Mean', 'Std_Dev',
       'num_checkpoints'],
      dtype='object')

# Import libraries
import pandas as pd
from sklearn.preprocessing import OneHotEncoder
from sklearn.decomposition import PCA
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt

# Create a dummy dataframe with both categorical and numerical columns

columns_to_drop=["Name","Description","RescuerID","PetID"]
categorical_columns=["Type",'Breed1', 'Breed2', 'Gender', 'Color1', 'Color2',
       'Color3', 'MaturitySize', 'FurLength', 'Vaccinated', 'Dewormed',
       'Sterilized', 'Health',"State"]
numerical_columns=['SentimentScore', 'Energy', 'Contrast', 'Correlation',
       'Entropy', 'Mean', 'Std_Dev', 'num_checkpoints','PhotoAmt','VideoAmt',"Age","Quantity","Fee"]

target=all_data["AdoptionSpeed"]
all_data=all_data.drop("AdoptionSpeed", axis=1)



#simple filling
all_data[categorical_columns] = all_data[categorical_columns].fillna("unknown")
all_data.fillna(all_data[numerical_columns].median(), inplace=True)
#all_data=all_data.fillna(0)

#transformation
all_data[categorical_columns]=all_data[categorical_columns].astype(str)
all_data[numerical_columns]=all_data[numerical_columns].astype(float)

all_data=all_data.drop(columns_to_drop, axis=1)

#normalize numerical columns:

for column in numerical_columns:
  all_data[numerical_columns] = (all_data[numerical_columns] - all_data[numerical_columns].mean()) / all_data[numerical_columns].std()


# Perform one-hot encoding on the categorical columns
encoder = OneHotEncoder(sparse=False)
encoded_cat = encoder.fit_transform(all_data[categorical_columns])
encoded_cat = pd.DataFrame(encoded_cat, columns=encoder.get_feature_names_out(categorical_columns))

# Concatenate the encoded categorical columns and the numerical columns
X = pd.concat([encoded_cat, all_data[numerical_columns]], axis=1)

# Perform PCA on the combined data
pca = PCA(n_components=14, )
pca_data = pca.fit_transform(X)

C:\Users\leoko\AppData\Roaming\Python\Python312\site-packages\sklearn\preprocessing\_encoders.py:975: FutureWarning: `sparse` was renamed to `sparse_output` in version 1.2 and will be removed in 1.4. `sparse_output` is ignored unless you leave `sparse` to its default value.
  warnings.warn(

pca_data = pd.DataFrame(pca_data)
print(pca.explained_variance_ratio_)

print(pca.components_)

[0.12700286 0.08180321 0.07040397 0.06507758 0.06149999 0.0502052
 0.04831881 0.04573533 0.04447697 0.03910402 0.03444903 0.02781854
 0.02519955 0.02393505]
[[-7.37783476e-02  7.37783476e-02 -4.39378707e-05 ... -3.15649782e-02
  -2.04128042e-02  6.60300265e-03]
 [ 6.94767437e-02 -6.94767437e-02  3.18644895e-05 ...  4.42598464e-01
  -5.28241794e-01  2.19804902e-01]
 [ 3.59065662e-02 -3.59065662e-02  9.08049614e-05 ... -4.37600209e-02
   1.63769363e-01  2.53406593e-04]
 ...
 [ 6.38520140e-02 -6.38520140e-02 -3.61414298e-05 ... -9.54149821e-02
   3.62494922e-02  3.09462182e-02]
 [ 7.90404995e-02 -7.90404995e-02  1.78586665e-05 ...  2.70340434e-01
  -9.83529011e-02  3.17846262e-02]
 [ 2.11774615e-02 -2.11774615e-02 -1.55401645e-04 ... -1.38743747e-02
   2.85332714e-01  5.94230316e-02]]

from itertools import combinations
plt.figure(figsize=(15, 12))
plt.subplots_adjust(hspace=0.5)
plt.suptitle("PCA plots with feature engineering")
n=1
for col1, col2 in combinations(pca_data.columns[:4], 2):
    ax = plt.subplot(3, 2, n)
    
    ax.set_xlabel(col1)
    ax.set_ylabel(col2)
    ax.set_title("pca{} vs pca{}".format(col1, col2)) 
    plt.scatter(pca_data[col1], pca_data[col2],c=target, label=target.unique(),cmap='rainbow',
  alpha=0.7,edgecolors='b')
    #plt.show()
    n=n+1

variance_exp_cumsum = pca.explained_variance_ratio_.cumsum().round(2)
fig, axes = plt.subplots(1,1,figsize=(16,7), dpi=100)
plt.plot(variance_exp_cumsum, color='firebrick')
plt.title('Screeplot of Variance Explained % with feature engineering', fontsize=22)
plt.xlabel('# of PCs', fontsize=16)
plt.show()

Model

import os
import json
import matplotlib.pyplot as plt
import seaborn as sns
import pprint
import warnings
import numpy as np
import pandas as pd

from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.decomposition import TruncatedSVD
from joblib import Parallel, delayed
from tqdm import tqdm, tqdm_notebook
from sklearn.model_selection import GridSearchCV
import os
from sklearn.model_selection import StratifiedKFold


# part 2
# required packages
from sklearn.compose import ColumnTransformer
from sklearn.pipeline import Pipeline
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import StandardScaler, OneHotEncoder

from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score, classification_report, confusion_matrix
from sklearn.svm import SVC

from textblob import TextBlob
import numpy as np # linear algebra
import pandas as pd
pd.set_option('display.max_rows', None)  # Display all rows
pd.set_option('display.max_columns', None)  # Display all columns
import json
import cv2
import skimage.feature as skf
import skimage.filters as skg
from skimage.measure import shannon_entropy
from scipy.stats import entropy
from sklearn.cluster import KMeans

from sklearn.model_selection import learning_curve
from sklearn.metrics import log_loss

import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
from torchvision import transforms
import pandas as pd

from sklearn.model_selection import train_test_split
from lime.lime_tabular import LimeTabularExplainer


%matplotlib inline

np.random.seed(seed=555)
warnings.filterwarnings('ignore')

# Load data
train = pd.read_csv('input/train/train.csv')
test = pd.read_csv('input/test/test.csv')
breed_labels = pd.read_csv('input/breed_labels.csv')
color_labels = pd.read_csv('input/color_labels.csv')
state_labels = pd.read_csv('input/state_labels.csv')

train.head()

	Type	Name	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	FurLength	Vaccinated	Dewormed	Sterilized	Health	Quantity	Fee	State	RescuerID	VideoAmt	Description	PetID	PhotoAmt	AdoptionSpeed
0	2	Nibble	3	299	0	1	1	7	0	1	1	2	2	2	1	1	100	41326	8480853f516546f6cf33aa88cd76c379	0	Nibble is a 3+ month old ball of cuteness. He ...	86e1089a3	1.0	2
1	2	No Name Yet	1	265	0	1	1	2	0	2	2	3	3	3	1	1	0	41401	3082c7125d8fb66f7dd4bff4192c8b14	0	I just found it alone yesterday near my apartm...	6296e909a	2.0	0
2	1	Brisco	1	307	0	1	2	7	0	2	2	1	1	2	1	1	0	41326	fa90fa5b1ee11c86938398b60abc32cb	0	Their pregnant mother was dumped by her irresp...	3422e4906	7.0	3
3	1	Miko	4	307	0	2	1	2	0	2	1	1	1	2	1	1	150	41401	9238e4f44c71a75282e62f7136c6b240	0	Good guard dog, very alert, active, obedience ...	5842f1ff5	8.0	2
4	1	Hunter	1	307	0	1	1	0	0	2	1	2	2	2	1	1	0	41326	95481e953f8aed9ec3d16fc4509537e8	0	This handsome yet cute boy is up for adoption....	850a43f90	3.0	2

# Display basic information about the data
train.info()

# Check for missing values
train.isnull().sum()

# Descriptive statistics
train.describe()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 14993 entries, 0 to 14992
Data columns (total 24 columns):
 #   Column         Non-Null Count  Dtype  
---  ------         --------------  -----  
 0   Type           14993 non-null  int64  
 1   Name           13728 non-null  object 
 2   Age            14993 non-null  int64  
 3   Breed1         14993 non-null  int64  
 4   Breed2         14993 non-null  int64  
 5   Gender         14993 non-null  int64  
 6   Color1         14993 non-null  int64  
 7   Color2         14993 non-null  int64  
 8   Color3         14993 non-null  int64  
 9   MaturitySize   14993 non-null  int64  
 10  FurLength      14993 non-null  int64  
 11  Vaccinated     14993 non-null  int64  
 12  Dewormed       14993 non-null  int64  
 13  Sterilized     14993 non-null  int64  
 14  Health         14993 non-null  int64  
 15  Quantity       14993 non-null  int64  
 16  Fee            14993 non-null  int64  
 17  State          14993 non-null  int64  
 18  RescuerID      14993 non-null  object 
 19  VideoAmt       14993 non-null  int64  
 20  Description    14980 non-null  object 
 21  PetID          14993 non-null  object 
 22  PhotoAmt       14993 non-null  float64
 23  AdoptionSpeed  14993 non-null  int64  
dtypes: float64(1), int64(19), object(4)
memory usage: 2.7+ MB

	Type	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	FurLength	Vaccinated	Dewormed	Sterilized	Health	Quantity	Fee	State	VideoAmt	PhotoAmt	AdoptionSpeed
count	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000	14993.000000
mean	1.457614	10.452078	265.272594	74.009738	1.776162	2.234176	3.222837	1.882012	1.862002	1.467485	1.731208	1.558727	1.914227	1.036617	1.576069	21.259988	41346.028347	0.056760	3.889215	2.516441
std	0.498217	18.155790	60.056818	123.011575	0.681592	1.745225	2.742562	2.984086	0.547959	0.599070	0.667649	0.695817	0.566172	0.199535	1.472477	78.414548	32.444153	0.346185	3.487810	1.177265
min	1.000000	0.000000	0.000000	0.000000	1.000000	1.000000	0.000000	0.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	0.000000	41324.000000	0.000000	0.000000	0.000000
25%	1.000000	2.000000	265.000000	0.000000	1.000000	1.000000	0.000000	0.000000	2.000000	1.000000	1.000000	1.000000	2.000000	1.000000	1.000000	0.000000	41326.000000	0.000000	2.000000	2.000000
50%	1.000000	3.000000	266.000000	0.000000	2.000000	2.000000	2.000000	0.000000	2.000000	1.000000	2.000000	1.000000	2.000000	1.000000	1.000000	0.000000	41326.000000	0.000000	3.000000	2.000000
75%	2.000000	12.000000	307.000000	179.000000	2.000000	3.000000	6.000000	5.000000	2.000000	2.000000	2.000000	2.000000	2.000000	1.000000	1.000000	0.000000	41401.000000	0.000000	5.000000	4.000000
max	2.000000	255.000000	307.000000	307.000000	3.000000	7.000000	7.000000	7.000000	4.000000	3.000000	3.000000	3.000000	3.000000	3.000000	20.000000	3000.000000	41415.000000	8.000000	30.000000	4.000000

# Visualize AdoptionSpeed distribution
sns.countplot(x='AdoptionSpeed', data=train)
plt.title('Adoption Speed Distribution')
plt.show()

Age

data = train.copy()

data_4 = data[data['AdoptionSpeed']==4]
data_rest = data[data['AdoptionSpeed']<4]
# Create subplots
plt.figure(figsize=(12, 6))
# Plot age distribution for fraud cases
plt.subplot(1, 2, 1)
sns.histplot(data=data_4, x='Age', kde=True, color='red')
plt.title("Age Distribution for not getting adopted")

# Plot age distribution for non-fraud cases
plt.subplot(1, 2, 2)
sns.histplot(data=data_rest, x='Age', kde=True, color='blue')
plt.title("Age Distribution for getting adopted")

# Show the plots
plt.tight_layout()
plt.show()

# checking age where month is greater than 100. 
np.where(train['Age']>100)

(array([  113,   145,   513,  1057,  1141,  1363,  1703,  1858,  1867,
         2263,  2497,  2767,  2830,  3195,  3684,  3931,  3998,  4071,
         4344,  4456,  4522,  4617,  4796,  5007,  5067,  5160,  5485,
         5535,  5569,  5593,  6264,  6569,  6618,  6666,  6722,  6894,
         6992,  7166,  7250,  7289,  7482,  7512,  7614,  7823,  8058,
         8065,  8304,  8492,  8944,  9267,  9378,  9563,  9970, 10089,
        10128, 10337, 10426, 10501, 10658, 10737, 10917, 11087, 11172,
        11216, 11314, 11523, 11605, 11640, 11968, 12526, 12575, 12621,
        12737, 12855, 12947, 13079, 13398, 13493, 13665, 13744, 14189,
        14303, 14637, 14732], dtype=int64),)

def age_segment(df):

    # Creating the "age_group" column by binning them into 4 groups
    conditions = [
        df['Age'] < 20,
        (df['Age'] >= 20) & (df['Age'] <= 40),
        (df['Age'] >= 40) & (df['Age'] <= 60),
        (df['Age'] >= 60) & (df['Age'] <= 80),
        (df['Age'] >= 80) & (df['Age'] <= 100)
    ]
    choices = ['1-20', '20-40', '40-60', '60-80','80-100']
    df['AgeGroup'] = np.select(conditions, choices, default='Unknown')
    
    return df

def extract_texture_features(image_path):
    # Load the image
    img = cv2.imread(image_path)

    # Convert the image to grayscale
    img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    
    haralick = skf.graycomatrix(img_gray, [1], [0], 256, symmetric=True, normed=True) 
    energy = skf.graycoprops(haralick, 'energy')[0, 0] 
    contrast = skf.graycoprops(haralick, 'contrast')[0, 0] 
    correlation = skf.graycoprops(haralick, 'correlation')[0, 0] 
    entropy = shannon_entropy(img_gray)

    mean = np.mean(img)
    std_dev = np.std(img)

    return energy, contrast, correlation, entropy, mean, std_dev

def extract_keypoints(file_path):
    # Load an image
    image = cv2.imread(file_path, cv2.IMREAD_GRAYSCALE)

    # Initialize the SIFT detector
    sift = cv2.SIFT_create()

    # Detect and compute SIFT keypoints and descriptors
    keypoints, _ = sift.detectAndCompute(image, None)

    # Return the number of keypoints
    return len(keypoints)


df_train_img['num_checkpoints'] = None  # Initialize the column

def extract_and_update(row):
    file_path = os.path.join(train_img_path, f"{row['PetID']}.jpg")
    num_keypoints = extract_keypoints(file_path)
    return num_keypoints


# Update the 'num_checkpoints' column
df_train_img['num_checkpoints'] = df_train_img.apply(extract_and_update, axis=1)
# df_train_img.to_csv('train_img_keypoints.csv')

# Columns to drop
columns_to_drop = ['mean_h', 'mean_s', 'mean_v', 'variance_h', 'variance_s', 'variance_v', 'Unnamed: 0']

# Drop specified columns
final_keypoints = df_train_img.drop(columns=columns_to_drop, errors='ignore')

# Print the resulting DataFrame
final_keypoints.head()

	PetID	Energy	Contrast	Correlation	Entropy	Mean	Std_Dev	num_checkpoints
0	0008c5398-1	0.024081	78.820303	0.989859	7.601080	143.223922	70.638848	319
0	0008c5398-2	0.070913	41.355089	0.994805	7.284530	91.658592	64.059420	233
0	0008c5398-3	0.040957	66.469614	0.989986	7.594550	111.068160	73.856578	410
0	0008c5398-4	0.050778	150.613044	0.984973	7.517500	108.980235	71.533391	469
0	0008c5398-5	0.025575	79.317245	0.992683	7.719211	141.169160	76.906123	395

cols = ['AdoptionSpeed','Description_Length','HasName','Age', 'Breed1', 'Breed2', 'MaturitySize', 'FurLength', 'Vaccinated',
       'Dewormed', 'Sterilized', 'Health', 'Quantity', 'Fee', 'State',
       'VideoAmt', 'PhotoAmt', 'Description_Length', 'HasName', 'Type_1',
       'Type_2', 'Gender_1', 'Gender_2', 'Gender_3']
corr_df = train[cols[:10]]
# # Explore correlations
correlation_matrix = corr_df.corr()
sns.heatmap(correlation_matrix, annot=True, cmap='coolwarm')
plt.title('Correlation Matrix')
plt.show()

# checking our new features
train.head()

	Type	Name	Age	Breed1	Breed2	Gender	Color1	Color2	Color3	MaturitySize	FurLength	Vaccinated	Dewormed	Sterilized	Health	Quantity	Fee	State	RescuerID	VideoAmt	Description	PetID	PhotoAmt	AdoptionSpeed	Description_Length	HasName	polarity	subjectivity
0	2	Nibble	3	299	0	1	1	7	0	1	1	2	2	2	1	1	100	41326	8480853f516546f6cf33aa88cd76c379	0	Nibble is a 3+ month old ball of cuteness. He ...	86e1089a3	1.0	2	359	1	0.130357	0.407143
1	2	No Name Yet	1	265	0	1	1	2	0	2	2	3	3	3	1	1	0	41401	3082c7125d8fb66f7dd4bff4192c8b14	0	I just found it alone yesterday near my apartm...	6296e909a	2.0	0	118	1	0.100000	0.400000
2	1	Brisco	1	307	0	1	2	7	0	2	2	1	1	2	1	1	0	41326	fa90fa5b1ee11c86938398b60abc32cb	0	Their pregnant mother was dumped by her irresp...	3422e4906	7.0	3	393	1	0.261905	0.542857
3	1	Miko	4	307	0	2	1	2	0	2	1	1	1	2	1	1	150	41401	9238e4f44c71a75282e62f7136c6b240	0	Good guard dog, very alert, active, obedience ...	5842f1ff5	8.0	2	146	1	0.341146	0.437500
4	1	Hunter	1	307	0	1	1	0	0	2	1	2	2	2	1	1	0	41326	95481e953f8aed9ec3d16fc4509537e8	0	This handsome yet cute boy is up for adoption....	850a43f90	3.0	2	390	1	0.400000	0.743750

F V 2.5

# Load data
train = pd.read_csv('input/train/train.csv')
test = pd.read_csv('input/test/test.csv')
breed_labels = pd.read_csv('input/breed_labels.csv')
color_labels = pd.read_csv('input/color_labels.csv')
state_labels = pd.read_csv('input/state_labels.csv')
df_image = pd.read_csv("image_features.csv", header = 0)

def get_polarity(text):
    return TextBlob(text).sentiment.polarity

def get_subjectivity(text):
    return TextBlob(text).sentiment.subjectivity

def get_features(train):
    # Creating a new feature based on the length of the description
    train['Description_Length'] = train['Description'].apply(lambda x: len(str(x)))
    # Creating a feature indicating whether a pet has a name or not
    train['HasName'] = train['Name'].apply(lambda x: 0 if pd.isnull(x) else 1)


    train['Polarity'] = train['Description'].astype(str).apply(get_polarity)
    train['Subjectivity'] = train['Description'].astype(str).apply(get_subjectivity)

    # Adding features engineering for train dataset
    to_scale = ['AgeScaled']
    scale_num = ['Age']
    train[to_scale] = (train[scale_num] - train[scale_num].mean()) / train[scale_num].std()

    # Concatenating Breed1, Age and Furlength features
    train['Breed1AgeFurLength'] = (train['Breed1'].astype(str) + np.abs(train['AgeScaled']).astype(str) + train['FurLength'].astype(str)).astype(float)
    train['Breed2AgeFurLength'] = (train['Breed2'].astype(str) + np.abs(train['AgeScaled']).astype(str) + train['FurLength'].astype(str)).astype(float)

    # Concatenating binary features such as Vaccinated, Dewormed and Sterilized
    train['VDSCombination'] = (train['Vaccinated'].astype(str) + train['Dewormed'].astype(str) + train['Sterilized'].astype(str)).astype(float)

    # Creating a color count feature without including cases where any color is 0
    train['ColorCount'] = train[['Color1', 'Color2', 'Color3']].apply(lambda row: len([color for color in row if color != 0]), axis=1)

    # Creating a total visual media feature
    train['TotalVisualMedia'] = train['PhotoAmt'] + train['VideoAmt']

    # Creating a description length feature
    train['DescriptionLength'] = train['Description'].apply(lambda x: len(str(x)))

    # Creating a sentiment score feature
    train['SentimentScore'] = train['Description'].apply(lambda x: TextBlob(str(x)).sentiment.polarity)
    
    #Name_word_len, 'RescuerID_3', 'RescuerID_5', 'Description_len', 'PhotoAmt', 'Name_word_len', 'is_emoji'
    train['Name_word_len'] = train['Name'].fillna('').apply(lambda x : len(x.split(' ')))

    to_drop = ['AgeScaled']
    train = train.drop(to_drop, axis = 1)
    
    # Adding image features to dataset

    # First finding the position of '-' in PetID
    df_image['pos'] = df_image['PetID'].str.find('-')
    df_image.drop(columns=['Unnamed: 0'], axis=1, inplace=True)

    # Using position to slice PetID using a lambda function
    df_image['PetID'] = df_image.apply(lambda x: x['PetID'][0:x['pos']],axis=1)
    
    df_image.drop(columns=['pos'], axis=1, inplace=True)

    train = pd.merge(train, df_image, how="inner", on=["PetID"], copy=True)

    return train
train = get_features(train)

class FEDenseNetP(nn.Module):
    def __init__(self, num_channels=3):
        super().__init__()
        preloaded = torchvision.models.densenet121(pretrained=True)
        self.features = preloaded.features
        self.features.conv0 = nn.Conv2d(num_channels, 64, kernel_size=8, stride=2, padding=3)
        self.average_pool = nn.AvgPool1d(4)
        
        del preloaded

    def forward(self, x):
        features = self.features(x)
        out = F.adaptive_avg_pool2d(features, (1, 1)).view(features.size(0), -1)
        x = self.average_pool(out)
        return x

transformer = transforms.Compose([        # Defining a variable transforms
                    transforms.Resize(256),                # Resize the image to 256×256 pixels
                    transforms.CenterCrop(256),            # Centre Cropping the image to 256×256 pixels about the center
                    transforms.ToTensor(),                 # Convert the image to PyTorch Tensor data type
                    transforms.Normalize(                  # Normalize the image
                    mean=[0.485, 0.456, 0.406],            # Mean and std of image as also used when training the network
                    std=[0.229, 0.224, 0.225]      
                    )])

def get_image_features(img,id='pet_id'):
    
    m = FEDenseNetP()
    m.eval()
    # loadin from PIL requires transformations
    try : 
#         img_t = transformer(img)
        img_t = torch.unsqueeze(img_t,0)
    except:
        img_t = img
        pass
    # fit the image to get the new weights
    y = m(img)
    # reduce dimension
    y = torch.squeeze(y)
    arr = y.data.numpy()
    return arr



   
    

import concurrent.futures
from tqdm import tqdm
from PIL import Image

features = {}
pet_ids = train['PetID'].values
path = 'data/train_images/'

# function to process a single image and return its features
def process_image(pet_id):
    image = Image.open(path + str(pet_id) + '-1.jpg').convert('RGB')
    input_tensor = transformer(image)
    input_batch = input_tensor.unsqueeze(0)
    features = get_image_features(input_batch)
    return pet_id, features

# parallel processing
with concurrent.futures.ThreadPoolExecutor() as executor:
    # Use tqdm to display progress
    with tqdm(total=len(pet_ids), desc="Processing Images", unit="image") as progress_bar:
        # Submit tasks to the executor and retrieve the results
        futures = [executor.submit(process_image, pet_id) for pet_id in pet_ids]
        for future in concurrent.futures.as_completed(futures):
            # Get the result and update the features dictionary
            pet_id, img_feats = future.result()
            features[pet_id] = img_feats
            progress_bar.update(1)
            
train_feats = pd.DataFrame.from_dict(features, orient='index')
train_feats.columns = [f'pic_{i}' for i in range(train_feats.shape[1])]
train_feats = train_feats.reset_index()
train_feats.rename({'index': 'PetID'}, axis='columns', inplace=True)

Processing Images: 100%|████████████████████████████████████████████████████| 58311/58311 [3:02:32<00:00,  5.32image/s]

n_components = 32
svd_ = TruncatedSVD(n_components=n_components, random_state=1337)

# features_df = pd.concat([train_feats, test_feats], axis=0)
features = train_feats[[f'pic_{i}' for i in range(256)]].values

svd_col = svd_.fit_transform(features)
svd_col = pd.DataFrame(svd_col)
svd_col = svd_col.add_prefix('IMG_SVD_')

all_ids = train['PetID']#pd.concat([train, test], axis=0, ignore_index=True, sort=False)[['PetID']]
img_features = pd.concat([all_ids, svd_col], axis=1)
# img_features.to_csv('img_features_v2_densenet121.csv')

train = train.merge(img_features, how='left', on='PetID')

img_feat2 = pd.read_csv("img_features_densenet121.csv", index_col=0)

train.shape

(409159, 75)

train.columns

Index(['Type', 'Name', 'Age', 'Breed1', 'Breed2', 'Gender', 'Color1', 'Color2',
       'Color3', 'MaturitySize', 'FurLength', 'Vaccinated', 'Dewormed',
       'Sterilized', 'Health', 'Quantity', 'Fee', 'State', 'RescuerID',
       'VideoAmt', 'Description', 'PetID', 'PhotoAmt', 'AdoptionSpeed',
       'Description_Length', 'HasName', 'Polarity', 'Subjectivity',
       'Breed1AgeFurLength', 'Breed2AgeFurLength', 'VDSCombination',
       'ColorCount', 'TotalVisualMedia', 'DescriptionLength', 'SentimentScore',
       'Name_word_len', 'Energy', 'Contrast', 'Correlation', 'Entropy', 'Mean',
       'Std_Dev', 'num_checkpoints', 'IMG_SVD_0', 'IMG_SVD_1', 'IMG_SVD_2',
       'IMG_SVD_3', 'IMG_SVD_4', 'IMG_SVD_5', 'IMG_SVD_6', 'IMG_SVD_7',
       'IMG_SVD_8', 'IMG_SVD_9', 'IMG_SVD_10', 'IMG_SVD_11', 'IMG_SVD_12',
       'IMG_SVD_13', 'IMG_SVD_14', 'IMG_SVD_15', 'IMG_SVD_16', 'IMG_SVD_17',
       'IMG_SVD_18', 'IMG_SVD_19', 'IMG_SVD_20', 'IMG_SVD_21', 'IMG_SVD_22',
       'IMG_SVD_23', 'IMG_SVD_24', 'IMG_SVD_25', 'IMG_SVD_26', 'IMG_SVD_27',
       'IMG_SVD_28', 'IMG_SVD_29', 'IMG_SVD_30', 'IMG_SVD_31'],
      dtype='object')

def get_estimators(train):
    cols_to_drop = ['PetID', 'AdoptionSpeed','Description','Name','RescuerID']
    X = train.drop(cols_to_drop, axis=1)
    y = train['AdoptionSpeed']
    Xtrain, Xtest, ytrain, ytest = train_test_split(X, y, test_size=0.2, random_state=42)
    model = RandomForestClassifier(n_jobs=-1)

    # intializing estimators
    estimators = np.arange(10, 300, 10)
    scores = []
    for n in estimators:
        model.set_params(n_estimators=n)
        model.fit(Xtrain, ytrain)
        scores.append(model.score(Xtest, ytest))

    #plotting the graph
    plt.title("Estimator for random forest")
    plt.xlabel("n_estimator")
    plt.ylabel("score")
    plt.plot(estimators, scores)

get_estimators(train)

From above we can either go with 200 parameters for a little bit of score improvement or choose 100 parameters for compute optimization as the score does not vary much.

With Encoding

def rf_we(train):
    numeric_features = ['Age','Fee','VideoAmt','PhotoAmt','Quantity']

    categorical_features = ['Type', 'Color1', 'Color2', 'Gender', 'MaturitySize', 
                        'FurLength', 'Vaccinated', 'Sterilized', 'Health', 'Breed1','Gender']

    numeric_transformer = Pipeline(steps=[
        ('scaler', StandardScaler())])


    categorical_transformer = Pipeline(steps=[

        ('onehot', OneHotEncoder(handle_unknown='ignore'))])

    preprocessor = ColumnTransformer(
        transformers=[
            ('num', numeric_transformer, numeric_features),
            ('cat', categorical_transformer, categorical_features)])


    cols_to_drop = ['PetID', 'AdoptionSpeed','Description','Name','RescuerID']
    X = train.drop(cols_to_drop, axis=1)
    y = train['AdoptionSpeed']
    X_train, X_valid, y_train, y_valid = train_test_split(X, y, test_size=0.2, random_state=42)


    model_1 = Pipeline(steps=[('preprocessor', preprocessor),
                      ('classifier', 
                       RandomForestClassifier(n_jobs=-1, n_estimators=200)
                         )])  
    model_1.fit(X_train,y_train)

    # make predictions on test data
    y_model_randomforest = model_1.predict(X_valid)

    # accuracy score
    train_accuracy_rf = model_1.score(X_train, y_train)
    test_accuracy_rf = model_1.score(X_valid, y_valid)
    print("Random Forest train accuracy: {:.2f}%".format(train_accuracy_rf * 100))
    print("Random Forest test accuracy: {:.2f}%".format(test_accuracy_rf * 100))
    return model_1

rf_we(train)

Random Forest train accuracy: 96.60%
Random Forest test accuracy: 90.61%

Pipeline(steps=[('preprocessor',
                 ColumnTransformer(transformers=[('num',
                                                  Pipeline(steps=[('scaler',
                                                                   StandardScaler())]),
                                                  ['Age', 'Fee', 'VideoAmt',
                                                   'PhotoAmt', 'Quantity']),
                                                 ('cat',
                                                  Pipeline(steps=[('onehot',
                                                                   OneHotEncoder(handle_unknown='ignore'))]),
                                                  ['Type', 'Color1', 'Color2',
                                                   'Gender', 'MaturitySize',
                                                   'FurLength', 'Vaccinated',
                                                   'Sterilized', 'Health',
                                                   'Breed1', 'Gender'])])),
                ('classifier',
                 RandomForestClassifier(n_estimators=200, n_jobs=-1))])

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

(58311, 75)

Since we have Class Imbalances lets use stratifiedkfold to evaluate

def rf_with_stratified_kfold(train, n_splits=5):
    numeric_features = ['Age','Fee','VideoAmt','PhotoAmt','Quantity']
    categorical_features = ['Type', 'Color1', 'Color2', 'Gender', 'MaturitySize', 
                            'FurLength', 'Vaccinated', 'Sterilized', 'Health', 'Breed1','Gender']

    numeric_transformer = Pipeline(steps=[('scaler', StandardScaler())])
    categorical_transformer = Pipeline(steps=[('onehot', OneHotEncoder(handle_unknown='ignore'))])

    preprocessor = ColumnTransformer(
        transformers=[
            ('num', numeric_transformer, numeric_features),
            ('cat', categorical_transformer, categorical_features)])

    cols_to_drop = ['PetID', 'AdoptionSpeed','Description','Name','RescuerID']
    X = train.drop(cols_to_drop, axis=1)
    y = train['AdoptionSpeed']

    # Initialize StratifiedKFold
    stratified_kfold = StratifiedKFold(n_splits=n_splits, shuffle=True, random_state=42)

    # Lists to store accuracy scores for each fold
    train_accuracies = []
    test_accuracies = []

    for train_index, test_index in stratified_kfold.split(X, y):
        X_train, X_valid = X.iloc[train_index], X.iloc[test_index]
        y_train, y_valid = y.iloc[train_index], y.iloc[test_index]

        model = Pipeline(steps=[
            ('preprocessor', preprocessor),
            ('classifier', RandomForestClassifier(n_jobs=-1, n_estimators=200))
        ])

        model.fit(X_train, y_train)

        # Making predictions on validation data
        y_model_randomforest = model.predict(X_valid)

        # Accuracy scores
        train_accuracy_rf = model.score(X_train, y_train)
        test_accuracy_rf = model.score(X_valid, y_valid)

        print("Random Forest train accuracy: {:.2f}%".format(train_accuracy_rf * 100))
        print("Random Forest test accuracy: {:.2f}%".format(test_accuracy_rf * 100))

        train_accuracies.append(train_accuracy_rf)
        test_accuracies.append(test_accuracy_rf)

    print("\nAverage Random Forest train accuracy: {:.2f}%".format(sum(train_accuracies) / n_splits * 100))
    print("Average Random Forest test accuracy: {:.2f}%".format(sum(test_accuracies) / n_splits * 100))

    return model

model_with_stratified_kfold = rf_with_stratified_kfold(train, n_splits=5)

Random Forest train accuracy: 98.01%
Random Forest test accuracy: 97.24%
Random Forest train accuracy: 98.02%
Random Forest test accuracy: 97.17%
Random Forest train accuracy: 98.01%
Random Forest test accuracy: 97.17%
Random Forest train accuracy: 98.00%
Random Forest test accuracy: 97.22%
Random Forest train accuracy: 98.02%
Random Forest test accuracy: 97.20%

Average Random Forest train accuracy: 98.01%
Average Random Forest test accuracy: 97.20%

model_with_stratified_kfold = rf_with_stratified_kfold(train, n_splits=5)

Random Forest train accuracy: 96.54%
Random Forest test accuracy: 91.06%
Random Forest train accuracy: 96.54%
Random Forest test accuracy: 91.41%
Random Forest train accuracy: 96.59%
Random Forest test accuracy: 90.94%
Random Forest train accuracy: 96.53%
Random Forest test accuracy: 91.05%
Random Forest train accuracy: 96.60%
Random Forest test accuracy: 90.94%

Average Random Forest train accuracy: 96.56%
Average Random Forest test accuracy: 91.08%

def plot_learning_curve_errors(model, train, title="Learning Curve"):
    cols_to_drop = ['PetID', 'AdoptionSpeed','Description','Name','RescuerID']
    X = train.drop(cols_to_drop, axis=1)
    y = train['AdoptionSpeed']
    train_sizes, train_errors, test_errors = learning_curve(
        model, X, y, cv=StratifiedKFold(n_splits=5, shuffle=True, random_state=42),
        train_sizes=[0.1, 0.3, 0.5, 0.7, 0.9], scoring='neg_log_loss', n_jobs=-1
    )

    train_errors_mean = -np.mean(train_errors, axis=1)
    train_errors_std = np.std(train_errors, axis=1)
    test_errors_mean = -np.mean(test_errors, axis=1)
    test_errors_std = np.std(test_errors, axis=1)

    plt.figure(figsize=(10, 6))
    plt.title(title)
    plt.xlabel("Training examples")
    plt.ylabel("Negative Log Loss")

    plt.fill_between(train_sizes, train_errors_mean - train_errors_std,
                     train_errors_mean + train_errors_std, alpha=0.1, color="r")
    plt.fill_between(train_sizes, test_errors_mean - test_errors_std,
                     test_errors_mean + test_errors_std, alpha=0.1, color="g")

    plt.plot(train_sizes, train_errors_mean, 'o-', color="r", label="Training error")
    plt.plot(train_sizes, test_errors_mean, 'o-', color="g", label="Validation error")

    plt.legend(loc="best")
    plt.show()

model = rf_we(train)
plot_learning_curve_errors(model, train, title="RF Learning Curve (Negative Log Loss)")

Random Forest train accuracy: 96.60%
Random Forest test accuracy: 90.66%

GridSearch CV

def gscv_we(train):   
    cols_to_drop = ['PetID', 'AdoptionSpeed','Description','Name','RescuerID']
    X = train.drop(cols_to_drop, axis=1)
    y = train['AdoptionSpeed']
    X_train, X_valid, y_train, y_valid = train_test_split(X, y, test_size=0.2, random_state=42)

    numerical_columns = train.select_dtypes(include=['number']).drop(['AdoptionSpeed'],axis = 1)
    numeric_features = list(numerical_columns.columns)
    categorical_features = ['Type', 'Color1', 'Color2', 'Gender', 'MaturitySize', 
                        'FurLength', 'Vaccinated', 'Sterilized', 'Health', 'Breed1','Gender']

    numeric_transformer = Pipeline(steps=[
        ('scaler', StandardScaler())])


    categorical_transformer = Pipeline(steps=[

        ('onehot', OneHotEncoder(handle_unknown='ignore'))])

    preprocessor = ColumnTransformer(
        transformers=[
            ('num', numeric_transformer, numeric_features),
            ('cat', categorical_transformer, categorical_features)])

    # Fitting the preprocessor on the training data
    preprocessor.fit(X_train)

    # pipeline for preprocessor and randomforest
    pipeline = Pipeline(steps=[('preprocessor', preprocessor),
                      ('classifier', 
                       RandomForestClassifier()
                         )])  



    # # Create the parameter grid based on the results of random search 
    param_grid = {
        'classifier__bootstrap': [True],  # Method of selecting samples for training each tree
        'classifier__max_depth': [120, 140, 150, 160],  # Maximum number of levels in tree
        'classifier__min_samples_leaf': [3, 4, 5],  # Minimum number of samples required at each leaf node
        'classifier__min_samples_split': [8, 10, 12],  # Minimum number of samples required to split a node
        'classifier__n_estimators': [100, 200, 300]  # Number of trees in random forest
    }

 


    # Instantiate the grid search model
    grid = GridSearchCV(estimator = pipeline, param_grid = param_grid, 
                              cv = 3, n_jobs = -1, verbose = 2)#scoring='accuracy'
    grid.fit(X_train, y_train);
    
    print(grid.best_params_)

    model = grid.best_estimator_
    y_model = model.predict(X_valid)
    y_model_train = model.predict(X_train)

    # accuracy score
    train_accuracy_rf = model.score(X_train, y_train)
    test_accuracy_rf = model.score(X_valid, y_valid)
    print("Random Forest train accuracy: {:.2f}%".format(train_accuracy_rf * 100))
    print("Random Forest test accuracy: {:.2f}%".format(test_accuracy_rf * 100))
    
    return model

%%time gscv_we(train) # fv2.5

Fitting 3 folds for each of 108 candidates, totalling 324 fits
Random Forest train accuracy: 98.52%
Random Forest test accuracy: 88.65%

model = rf_we(train)
columns = X.columns
plot_feature_importance(model.feature_importances_,columns[:],'RANDOM FOREST',10)

def lime_interpretations(train):
    cols_to_drop = ['PetID', 'AdoptionSpeed', 'Description', 'Name', 'RescuerID']
    X = train.drop(cols_to_drop, axis=1)
    y = train['AdoptionSpeed']


    # Split the data into training and testing sets
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

    # Train a simple model (e.g., RandomForestClassifier)
    model = RandomForestClassifier(n_estimators=200, random_state=42)
    model.fit(X_train, y_train)

    # Initialize LIME explainer
    explainer = LimeTabularExplainer(X_train.values, feature_names=X_train.columns)

    # Choose an instance from the test set for interpretation
    instance_to_explain = X_test.iloc[0]

    # Explain the instance using LIME
    exp = explainer.explain_instance(instance_to_explain.values, model.predict_proba, num_features=len(X_train.columns))

    # Display the explanation
    exp.show_in_notebook()

lime_interpretations(train)